HK1207124B - Non-invasive determination of methylome of fetus or tumor from plasma - Google Patents

Description

Non-invasive determination of fetal or tumor methylome from plasma

Cross-citation of related applications

This application is a PCT application claiming priority to the following: U.S. Provisional Patent Application No. 61/830,571, filed June 3, 2013, entitled "Tumor Detection In Plasma Using Methylation Status and Copy Number"; and U.S. Application No. 13/842, filed March 15, 2013, entitled "Non-Invasive Determination of Methylome of Fetus or Tumor From Plasma". Application No. 209 is a non-provisional application filed on September 20, 2012, entitled "Method of Determining the Whole Genome DNA Methylation Status of the Placenta by Massively Parallel Sequencing of Maternal Plasma," and claims the benefit of U.S. Provisional Patent Application No. 61/703,512, which is incorporated herein by reference in its entirety for all purposes.

Technical Field

This invention generally relates to the determination of DNA methylation patterns (methylome), and more specifically, to the analysis of biological samples (e.g., plasma) comprising a mixture of DNA from different genomes (e.g., from fetus and mother, or from tumor and normal cells) to determine the methylation patterns (methylome) of a minority of genomes. The uses of the determined methylome are also described.

Background Technology

Embryonic and fetal development is a complex process involving a series of highly coordinated genetic and epigenetic events. Cancer development is also a complex process, typically involving multiple genetic and epigenetic steps. Abnormalities in the epigenetic control of developmental processes are associated with infertility, miscarriage, intrauterine growth disorders, and postnatal outcomes. DNA methylation is one of the most frequently studied epigenetic mechanisms. DNA methylation mostly occurs when a methyl group is added to the 5' carbon of a cytosine residue in a CpG dinucleotide. Cytosine methylation adds a layer of control over gene transcription and DNA function. For example, hypermethylation of CpG-rich dinucleotide-rich gene promoters, known as CpG islands, is often associated with the repression of gene function.

Although epigenetic mechanisms play a crucial role in regulating development, human embryonic and fetal tissues are not readily available for analysis (tumors may be similarly unavailable). Studying the dynamic changes in such epigenetic processes of health and disease during the prenatal period in humans is practically impossible. Extraembryonic tissues, particularly the placenta, can be obtained as part of prenatal diagnostic procedures or postnatally, providing one of the main avenues for such research. However, these tissues require invasive procedures.

The DNA methylation patterns of the human placenta have fascinated researchers for decades. The human placenta exhibits a wealth of rare physiological features associated with DNA methylation. At the global level, placental tissue is hypomethylated compared to most somatic tissues. At the gene level, the methylation status of selected loci is a tissue-specific marker for placental tissue. Both global and locus-specific methylation patterns demonstrate gestational age-dependent alterations. Imprinted genes, allele-dependent parental genes, play a crucial role in the placenta. The placenta has been described as pseudomalignant, and hypermethylation of several tumor suppressor genes has been observed.

Studies of DNA methylation patterns in placental tissue have enabled insights into the pathophysiology of pregnancy-related or developmental disorders, such as preeclampsia and intrauterine growth restriction. Genomically imprinted disorders are associated with developmental disorders such as Prader-Willi syndrome and Angelman syndrome. Genomic imprinting and altered patterns of overall DNA methylation in placental and fetal tissue have been observed in pregnancies resulting from assisted reproductive technologies (H Hiura et al. 2012 HumReprod; 27:2541-2548). For example, a large number of environmental factors, such as maternal smoking (KE Haworth et al. 2013 Epigenomics; 5:37-49), maternal dietary factors (X Jiang et al. 2012 FASEB J; 26:3563-3574), and maternal metabolic status (e.g. diabetes) (N Hajj et al., Diabetes. doi:10.2337/db12-0289), are associated with epigenetic abnormalities in offspring.

Despite decades of effort, no truly feasible method exists for studying fetal or tumor methylomes and monitoring dynamic changes during pregnancy or disease processes such as malignancies. Therefore, providing a non-invasive method for analyzing all or part of the fetal methylome and tumor methylome is of great value.

Summary of the Invention

The examples provide systems, methods, and apparatus for determining and using methylation patterns in various tissues and samples. Illustrative examples are provided. The methylation pattern of fetal/tumor tissue can be inferred based on a comparison of plasma methylation (or other samples containing cell-free DNA, such as urine, saliva, genital wash) with the methylation pattern of the mother/patient. When the sample contains a mixture of DNA, tissue-specific alleles can be used to identify DNA from the fetus/tumor to determine the methylation pattern of the fetal/tumor tissue. Methylation pattern can be used to determine copy number variations in the fetal/tumor genome. Methylation markers in the fetus are identified using various techniques. The methylation pattern can be determined by measuring size parameters of the size distribution of DNA fragments, where reference values for the size parameters can be used to determine the methylation level.

In addition, methylation levels can be used to determine cancer grade. In cancer cases, measurements of changes in plasma methylome can allow for cancer detection (e.g., for screening purposes), monitoring (e.g., detecting response to anticancer therapy; and detecting cancer recurrence), and prognosis (e.g., for measuring the burden of cancer cells in the body, for cancer staging, or for assessing the probability of death due to the disease or its progression or metastasis).

The nature and advantages of the embodiments of the present invention can be better understood by referring to the following detailed implementation and accompanying drawings.

Attached Figure Description

Figure 1A shows Table 100, which represents the sequencing results of maternal blood, placenta, and maternal plasma according to an embodiment of the present invention.

Figure 1B shows the methylation density in a 1Mb window of a sequencing sample according to an embodiment of the present invention.

Figures 2A-2C show the curves of β values versus methylation index: (A) maternal blood cells, (B) chorionic villus sample, (c) full-term placental tissue.

Figures 3A and 3B show bar charts of the percentage of methylated CpG sites in plasma and blood cells collected from adult men and non-pregnant adult women: (A) autosomes, (B) chromosome X.

Figures 4A and 4B show the methylation density curves of corresponding loci in blood cell DNA and plasma DNA: (A) non-pregnant adult women, (B) adult men.

Figures 5A and 5B show bar charts of the percentage of methylated CpG sites in samples collected from pregnant women: (A) autosomes, (B) chromosome X.

Figure 6 shows a bar chart of methylation levels for different repetitive sequence classes in the human genome from maternal blood, placenta, and maternal plasma.

Figure 7A shows a Circos diagram 700 of an early pregnancy sample. Figure 7B shows a Circos diagram 750 of a late pregnancy sample.

Figures 8A-8D show comparative curves of methylation density of genomic tissue DNA versus maternal plasma DNA around CpG sites surrounding single nucleotide polymorphisms that provide useful information.

Figure 9 is a flowchart illustrating a method 900 for determining a first methylation morphology from a biological sample of an organism according to an embodiment of the present invention.

Figure 10 is a flowchart illustrating a method 1000 for determining a first methylation type from a biological sample of an organism according to an embodiment of the present invention.

Figures 11A and 11B show performance graphs of a prediction algorithm using maternal plasma data and fetal DNA percentage concentration according to an embodiment of the present invention.

Figure 12A is Table 1200, showing details of the predicted methylation of 15 selected loci according to embodiments of the present invention. Figure 12B is Graph 1250, showing the inferred categories of the 15 selected loci in the placenta and their corresponding methylation levels in the placenta.

Figure 13 is a flowchart of method 1300, which is used to detect fetal chromosomal abnormalities from a biological sample of a female individual carrying at least one fetus.

Figure 14 is a flowchart of method 1400, which is used to identify methylated molecular markers by comparing placental methylation patterns with maternal methylation patterns according to an embodiment of the present invention.

Figure 15A, Table 1500, demonstrates the performance of the DMR identification algorithm for early pregnancy data using 33 previously reported methylation molecular markers for early pregnancy. Figure 15B, Table 1550, demonstrates the performance of the DMR identification algorithm using late pregnancy data and comparing it with placental samples obtained at delivery.

Figure 16 is Table 1600, showing the number of loci predicted as hypermethylated or hypomethylated based on direct analysis of maternal plasma bisulfite sequencing data.

Figure 17A is curve 1700, showing the size distribution of DNA in maternal plasma, non-pregnant female control plasma, placenta, and peripheral blood. Figure 17B is curve 1750, showing the size distribution and methylation patterns in maternal plasma, adult female control plasma, placental tissue, and adult female control blood.

Figures 18A and 18B are graphs showing the methylation density and size of plasma DNA molecules according to embodiments of the present invention.

Figure 19A shows the methylation density and size curves for sequencing reads from adult non-pregnant women at 1900. Figure 19B is Figure 1950, showing the size distribution and methylation patterns of fetal-specific and maternal-specific DNA molecules in maternal plasma.

Figure 20 is a flowchart of method 2000, which is used to evaluate the methylation level of DNA in a biological sample of an organism according to an embodiment of the present invention.

Figure 21A is Table 2100, showing the methylation density of plasma and tissue samples from patients with hepatocellular carcinoma (HCC) before surgery. Figure 21B is Table 2150, showing the number of sequence reads and sequencing depth achieved for each sample.

Figure 22 is a copy of Table 220, showing the methylation density in autosomes in plasma samples from healthy controls, ranging from 71.2% to 72.5%.

Figures 23A and 23B show the methylation density of leukocyte layer, tumor tissue, non-tumor liver tissue, preoperative plasma, and postoperative plasma in HCC patients.

Figure 24A is Figure 2400, showing the methylation density of plasma from HCC patients before surgery. Figure 24B is curve 2450, showing the methylation density of plasma from HCC patients after surgery.

Figures 25A and 25B show the z-scores of plasma DNA methylation density in plasma samples from HCC patients before (Figure 2500) and after (Figure 2550) surgery, using plasma methylome data from four healthy control individuals as a reference, for chromosome 1.

Figure 26A, Table 2600, shows the z-score data for pre- and post-operative plasma. Figure 26B, Circos Figure 2620, shows the z-scores of plasma DNA methylation density in pre- and post-operative plasma samples from HCC patients, targeting 1Mb regions analyzed from all autosomes, using four healthy controls as references. Figure 26C, Table 2640, shows the distribution of z-scores for 1Mb regions across the entire genome in pre- and post-operative plasma samples from HCC patients. Figure 26D, Table 2660, shows the methylation levels in tumor tissue and pre-operative plasma samples overlapping with some control plasma samples, using CHH and CHG backgrounds.

Figures 27A-H show Circos plots of methylation density for eight cancer patients according to an embodiment of the present invention. Figure 27I is Table 2780, showing the number of sequence reads and sequencing depth achieved for each sample. Figure 27J is Table 2790, showing the distribution of z-scores for regions with a resolution of 1 Mb at the whole genome level in the plasma of patients with different malignant diseases. CL = Lung adenocarcinoma; NPC = Nasopharyngeal carcinoma; CRC = Colorectal cancer; NE = Neuroendocrine carcinoma; SMS = Leiomyosarcoma.

Figure 28 is a flowchart of method 2800, which, according to an embodiment of the present invention, analyzes a biological sample of an organism to determine the classification of cancer grade.

Figure 29A, curve 2900, shows the distribution of methylation density in the reference individual, assuming this distribution follows a normal distribution. Figure 29B, curve 2950, shows the distribution of methylation density in the cancer individual, assuming this distribution follows a normal distribution and the mean methylation level is 2 standard deviations below the threshold.

Figure 30 is curve 3000, showing the distribution of methylation density in plasma DNA of healthy individuals and cancer patients.

Figure 31 is Chart 3100, showing the distribution of the difference in methylation density between plasma DNA in healthy individuals and the mean methylation density between tumor tissues of HCC patients.

Figure 32A is Table 3200, showing the effect of reducing sequencing depth when plasma samples contain 5% or 2% tumor DNA.

Figure 32B is Figure 3250, showing the methylation density of repeating elements and non-repeat regions in plasma and leukocyte layer of HCC patients, normal liver tissue, tumor tissue, preoperative plasma and postoperative plasma samples from four healthy control individuals.

Figure 33 shows a block diagram of an exemplary computer system 3300 that can be used with systems and methods according to embodiments of the present invention.

Figure 34A shows the size distribution of plasma DNA in SLE04, a patient with systemic lupus erythematosus (SLE). Figures 34B and 34C show the methylation analysis of plasma DNA from SLE04 (Figure 34B) and TBR36 (Figure 34C), a patient with hepatocellular carcinoma (HCC).

Figure 35 is a flowchart of method 3500, which, according to an embodiment of the present invention, determines the classification of cancer grades based on the hypermethylation of CpG islands.

Figure 36 is a flowchart of method 3600, which, according to an embodiment of the present invention, analyzes biological samples of an organism using multiple chromosome regions.

Figure 37A shows the CNA analysis of tumor tissue, untreated (BS) plasma DNA, and BS-treated plasma DNA (from inside to outside) in patient TBR36. Figure 37B is a scatter plot showing the relationship between the z-scores of CNAs in the 1 Mb region detected using BS-treated and untreated plasma in patient TBR36.

Figure 38A shows the CNA analysis of tumor tissue, untreated (BS) plasma DNA, and BS-treated plasma DNA (from inside to outside) in patient TBR34. Figure 38B is a scatter plot showing the relationship between the z-scores of CNAs in the 1 Mb region detected using BS-treated and untreated plasma for patient TBR34.

Figure 39A is a Circos diagram showing the CNA (inner ring) and methylation (outer ring) analysis of bisulfite-treated plasma from HCC patient TBR240. Figure 39B is a Circos diagram showing the CNA (inner ring) and methylation (outer ring) analysis of bisulfite-treated plasma from HCC patient TBR164.

Figure 40A shows CNA analysis of pre- and post-treatment samples of patient TBR36. Figure 40B shows methylation analysis of pre- and post-treatment samples of patient TBR36. Figure 41A shows CNA analysis of pre- and post-treatment samples of patient TBR34. Figure 41B shows methylation analysis of pre- and post-treatment samples of patient TBR34.

Figure 42 shows the diagnostic performance of whole-genome hypomethylation analysis with different numbers of sequencing reads.

Figure 43 is a graph showing the ROC curves for detecting cancer based on whole-genome hypomethylation analysis using different region sizes (50kb, 100kb, 200kb and 1Mb).

Figure 44A shows the diagnostic performance of cumulative probability (CP) and the percentage of regions with abnormalities. Figure 44B shows the diagnostic performance of plasma analysis for overall hypomethylation, CpG island hypermethylation, and CNA.

Figure 45 shows a table of results for overall hypomethylation, CpG island hypermethylation, and CNA in patients with hepatocellular carcinoma.

Figure 46 shows a table of results for overall hypomethylation, CpG island hypermethylation, and CNA in patients with cancers other than hepatocellular carcinoma.

Figure 47 shows a series of analyses of plasma methylation in case TBR34.

Figure 48A shows a Circos plot confirming CNA (inner loop) and methylation alterations (outer loop) in bisulfite-treated plasma DNA of HCC patient TBR36. Figure 48B is a box plot of methylation z-fractions in regions with chromosomal increases and deletions, as well as regions without copy number alterations, of HCC patient TBR36.

Figure 49A shows a Circos diagram confirming CNA (inner loop) and methylation alterations (outer loop) in bisulfite-treated plasma DNA of TBR34 from HCC patients. Figure 49B is a box plot of methylation z-fractions in regions with chromosomal increases and deletions, as well as regions without copy number alterations, of TBR34 from HCC patients.

Figures 50A and 50B show the results of plasma hypomethylation and CNA analysis in SLE04 and SLE10 patients.

Figures 51A and 51B show Z- _methylation analysis of plasma regions with and without CNA in two HCC patients (TBR34 and TBR36). Figures 51C and 51D show Z- _methylation analysis of plasma regions with and without CNA in two SLE patients (SLE04 and SLE10).

Figure 52A shows the hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using Group A features, including CNA, global methylation, and CpG island methylation. Figure 52B shows the hierarchical clustering using Group B features, including CNA, global methylation, and CpG island methylation.

Figure 53A shows a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using group A CpG island methylation characteristics. Figure 53B shows a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using group A overall methylation density.

Figure 54A shows the hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using group A overall CNA. Figure 54B shows the hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using group B CpG island methylation density.

Figure 55A shows the hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using the overall methylation density of group B. Figure 55B shows the hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using the overall methylation density of group B.

Figure 56 shows the average methylation density (red dots) in the 1Mb region among 32 healthy individuals.

Detailed Implementation

definition

The “methylome” provides a measure of the amount of DNA methylation at multiple sites or loci in the genome. A methylome can correspond to the entire genome, a large portion of the genome, or a relatively small portion of the genome. The “fetal methylome” corresponds to the methylome of the fetus of a pregnant woman. The fetal methylome can be determined using a variety of fetal tissue or fetal DNA sources, including placental tissue and cell-free fetal DNA in maternal plasma. The “tumor methylome” corresponds to the methylome of a tumor in an organism (e.g., a human). The tumor methylome can be determined using tumor tissue or cell-free tumor DNA in maternal plasma. The fetal methylome and the tumor methylome are examples of related methylomes. Other examples of relevant methylomes include the methylomes of organs (e.g., brain cells, bone, lungs, heart, muscles, and kidneys) that can provide DNA into bodily fluids (e.g., plasma, serum, sweat, saliva, urine, genital secretions, semen, fecal fluid, diarrheal fluid, cerebrospinal fluid, gastrointestinal secretions, pancreatic secretions, intestinal secretions, sputum, tears, aspirated fluid from the breast and thyroid gland, etc.). The organ can be a transplanted organ.

"Plasma methylome" refers to the methylome measured from the plasma or serum of an animal (e.g., a human). Because plasma and serum contain cell-free DNA (cell-free DNA refers to DNA present outside cells), the plasma methylome is an example of the cell-free methylome. The plasma methylome is also an example of a mixed methylome, as it is a mixture of fetal/maternal methylomes or tumor/patient methylomes. "Placental methylome" can be measured from chorionic villus sampling (CVS) or placental tissue samples (e.g., postpartum). "Cellular methylome" corresponds to the methylome measured from the patient's cells (e.g., blood cells). The methylome of blood cells is called the blood cell methylome (or blood methylome).

A "site" corresponds to a single location, which can be a single base position or a group of related base positions, such as a CpG site. A "locus" can correspond to a region that includes multiple sites. A locus can also include only one site, which would make the locus equivalent to a single site in this context.

The “methylation index” for each genomic locus (e.g., a CpG site) refers to the proportion of sequence reads showing methylation at the site relative to the total number of reads covering that site. The “methylation density” of a region is the number of reads showing methylation at sites within that region divided by the total number of reads covering sites within that region. Sites can have specific characteristics, such as being CpG sites. Therefore, the “CpG methylation density” of a region is the number of reads showing CpG methylation divided by the total number of reads covering CpG sites (e.g., a specific CpG site, CpG sites within a CpG island, or a larger region) within that region. For example, the methylation density of each 100kb region in the human genome can be calculated as the proportion of the total number of unconverted cytosine (corresponding to methylated cytosine) at CpG sites after bisulfite treatment to all CpG sites measured and aligned to the 100kb region's sequence reads covering all CpG sites. This analysis can also be performed for other region sizes, such as 50kb or 1Mb. A region can be an entire genome or a chromosome or a portion of a chromosome (e.g., a chromosome arm). When a region contains only CpG sites, the methylation index of the CpG sites is the same as the methylation density of the region. The "ratio of methylated cytosine" refers to the number of methylated (e.g., unconverted after bisulfite conversion) cytosine sites "C" out of the total number of cytosine residues analyzed, i.e., cytosine included in the region outside the CpG background. The methylation index, methylation density, and ratio of methylated cytosine are examples of "methylation level."

"Methylation pattern" (also known as methylation state) includes information related to DNA methylation in a region. Information related to DNA methylation may include (but is not limited to) the methylation index of CpG sites, the methylation density of CpG sites in a region, the distribution of CpG sites in adjacent regions, the pattern or level of methylation at each individual CpG site in a region containing more than one CpG site, and non-CpG methylation. Most of the methylation pattern in the genome can be considered equivalent to the methylome. In mammalian genomes, "DNA methylation" typically refers to the addition of a methyl group to the 5' carbon of a cytosine residue in a CpG dinucleotide (i.e., 5-methylcytosine). DNA methylation can occur in cytosine in other backgrounds, such as CHG and CHH, where H is adenine, cytosine, or thymine. Cytosine methylation can also occur in the form of 5-hydroxymethylcytosine. Non-cytosine methylation, such as N6-methyladenine, has also been reported.

"Tissue" corresponds to any cell. Different types of tissue can correspond to different types of cells (e.g., liver, lung, or blood), but can also correspond to tissues or healthy cells and tumor cells from different organisms (mother and fetus). "Biological sample" refers to any sample taken from an individual (e.g., a pregnant woman, a cancer patient or suspected cancer patient, an organ transplant recipient or an individual suspected of having a disease process involving an organ (e.g., the heart in a myocardial infarction or the brain in a stroke)) and containing one or more relevant nucleic acid molecules. Biological samples can be bodily fluids, such as blood, plasma, serum, urine, vaginal fluid, uterine or vaginal lavage fluid, pleural fluid, ascites, cerebrospinal fluid, saliva, sweat, tears, sputum, bronchoalveolar lavage fluid, etc. Fecal samples may also be used.

The term "cancer grade" can refer to the presence of cancer, cancer stage, tumor size, presence of metastasis, total tumor burden in the body, and/or other measures of cancer severity. Cancer grade can be numerical or other characteristics. The grade can be zero. Cancer grade also includes precancerous or precancerous symptoms (states) associated with a mutation or multiple mutations. Cancer grade can be used in various ways. For example, screening can check whether someone who is known not to have cancer has cancer. Assessment can study someone who has been diagnosed with cancer to monitor cancer progression over time, investigate the efficacy of treatments, or determine prognosis. In one embodiment, prognosis can be expressed as the probability that a patient will die from cancer, or the probability that cancer will progress after a specific duration or time, or the probability that cancer will metastasize. Testing can mean 'screening' or it can mean checking whether someone has suggestive characteristics of cancer (such as symptoms or other positive tests) has cancer.

Invention Details

Epigenetic mechanisms play a crucial role in embryonic and fetal development. However, human embryos and fetal tissues (including placental tissue) are not readily available (US Patent 6,927,028). This problem has been addressed in some embodiments by analyzing samples containing cell-free fetal DNA molecules present in maternal circulation. The fetal methylome can be inferred in several ways. For example, the maternal plasma methylome can be compared with the cellular methylome (from the mother's blood cells) and show differences associated with the fetal methylome. As another example, fetal-specific alleles can be used to determine the methylation of the fetal methylome at specific loci. Furthermore, as demonstrated by the correlation between size and methylation percentage, fragment size can be used as an indicator of methylation percentage.

In one embodiment, whole-genome bisulfite sequencing is used to analyze the methylation patterns (partial or complete methylome) of maternal plasma DNA at single nucleotide resolution. By employing polymorphic differences between mother and fetus, the fetal methylome can be derived from the maternal blood sample. In another embodiment, instead of using polymorphic differences, the difference between the plasma methylome and the blood cell methylome can be used.

In another embodiment, tumor methylation pattern analysis can be performed on samples from patients suspected or known to have cancer by employing single nucleotide variants and/or copy number abnormalities between the tumor genome and non-tumor genomes, as well as sequencing data from plasma (or other samples). Differences in methylation levels in the plasma sample of a tested individual, when compared to plasma methylation levels in healthy controls or a group of healthy controls, can allow for the identification of the tested individual as having cancer. Furthermore, methylation pattern can serve as a marker revealing the type of cancer, such as which organ it originated from, whether the individual has developed cancer, and whether metastasis has occurred.

Because this method is non-invasive, it allows for continuous assessment of fetal and maternal plasma methylomes from maternal blood samples collected in early pregnancy, late pregnancy, and postpartum. Pregnancy-related changes were observed. The method can also be applied to samples obtained during mid-pregnancy. The fetal methylome inferred from maternal plasma during pregnancy is similar to the placental methylome. Regions with imprinted genes and differential methylation were identified from maternal plasma data.

Therefore, a non-invasive, continuous, and comprehensive method for studying the fetal methylome has been developed, enabling the identification of biomarkers or direct testing for pregnancy-related lesions. Examples can also be used for non-invasive, continuous, and comprehensive studies of the tumor methylome for screening or detecting cancer in individuals, monitoring malignancies in cancer patients, and for prognosis. Examples can be applied to any type of cancer, including (but not limited to) lung cancer, breast cancer, colorectal cancer, prostate cancer, nasopharyngeal carcinoma, gastric cancer, testicular cancer, skin cancer (e.g., melanoma), cancers affecting the nervous system, bone cancer, ovarian cancer, liver cancer (e.g., hepatocellular carcinoma), hematological malignancies, pancreatic cancer, endometrial cancer, kidney cancer, cervical cancer, bladder cancer, etc.

First, a description of how to determine methylomes or methylation patterns is presented. Then, different methylomes are described (e.g., fetal methylome, tumor methylome, maternal or patient methylome, and mixed methylomes, such as those derived from plasma). The determination of fetal methylation patterns using fetal-specific markers or by comparing mixed methylation patterns with cellular methylation patterns is then described. Fetal methylation markers are determined by comparing methylation patterns. The relationship between size and methylation is discussed. The use of methylation patterns in cancer detection is also presented.

I. Determination of the methylation group

Numerous methods have been used to study the placental methylome, but each has its limitations. For example, sodium bisulfite, a chemical that modifies unmethylated cytosine residues into uracil while preserving methylated cytosine, translates differences in cytosine methylation into differences in gene sequence for further inquiry. The gold standard for studying cytosine methylation is based on treating tissue DNA with sodium bisulfite, followed by direct sequencing of individual clones of the bisulfite-converted DNA molecules. After analyzing multiple clones of the DNA molecule, the cytosine methylation pattern and quantitative morphology at each CpG site can be obtained. However, clonal bisulfite sequencing is a low-throughput and labor-intensive procedure, not easily applied on a whole-genome scale.

Methylation-sensitive restriction enzymes, which digest unmethylated DNA, typically provide a low-cost method for studying DNA methylation. However, the data generated by such studies are limited to loci with enzyme recognition motifs, and the results are not quantitative. Immunoprecipitation of DNA bound to anti-methylating cytosine antibodies can be used to study large regions of the genome, but tends to favor loci with dense methylation because antibodies bind more strongly to such regions. Microarray-based methods rely on the prior design of the probes used in the study and the hybridization efficiency between the probes and target DNA.

To comprehensively study the methylome, some embodiments use massively parallel sequencing (MPS) to provide genome-wide information and quantitative assessment of methylation levels based on each nucleotide and each allele. Recently, genome-wide MPS after pre-bisulfite transformation has become feasible (Lister et al. 2008 Cell; 133:523-536).

A small number of published studies have applied whole-genome bisulfite sequencing to the study of the human methylome (Lister et al. 2009 Nature; 462:315-322); Laurent et al. 2010 Genome Research; 20:320-331; Li et al. 2010 PRS; 8:e In *Lister et al.* 2009 *Nature* 462:315-322; and *Laurent et al.* 2010 *Genomics Research* 20:20-331, both studies focused on embryonic stem cells and fetal fibroblasts (Lister et al. 2009 *Nature* 462:315-322; Laurent et al. 2010 *Genomics Research* 20:320-331). Both studies analyzed cell line-derived DNA.

A. Whole-genome bisulfite sequencing

Certain embodiments address the aforementioned challenges and enable comprehensive, non-invasive, and continuous studies of the fetal methylome. In one embodiment, whole-genome bisulfite sequencing is used to analyze cell-free fetal DNA molecules found in the maternal bloodstream. Despite the low abundance and fragmentation of plasma DNA molecules, we are able to assemble a high-resolution fetal methylome from maternal plasma and observe continuous changes associated with pregnancy progression. Assuming a strong interest in non-invasive prenatal testing (NIPT), these embodiments could provide a powerful new tool for discovering fetal biomarkers or serve as a direct platform for NIPT of fetal or pregnancy-related conditions. Data from whole-genome bisulfite sequencing of various samples are currently available from which the fetal methylome can be derived. In one embodiment, this technique can be applied to the analysis of methylation patterns in pregnant women with preeclampsia, intrauterine growth restriction, or preterm labor. For such pregnancy complications, this technique can be used continuously due to its non-invasive nature, allowing for monitoring and/or prognosis and/or monitoring of response to treatment.

Figure 1A shows Table 100, illustrating the sequencing results of maternal blood, placenta, and maternal plasma according to an embodiment of the present invention. In one embodiment, whole-genome sequencing was performed on bisulfite-converted DNA libraries prepared using methylated DNA library adaptors (Illumina) (Lister et al. 2008 Cell; 133:523-536): blood cells from blood samples collected in early pregnancy, CVS, placental tissue collected at delivery, and maternal plasma samples collected during early and late pregnancy and the postpartum period. Blood cell and plasma DNA samples obtained from an adult male and an adult non-pregnant woman were also analyzed. In this study, a total of 9.5 billion pairs of raw sequence reads were generated. The sequencing coverage for each sample is shown in Table 100.

For maternal plasma samples from early pregnancy, late pregnancy, and postpartum, the sequence reads uniquely aligned to the human reference genome achieved average haploid genome coverage of 50-fold, 34-fold, and 28-fold, respectively. For samples obtained from pregnant women, CpG site coverage in the genome ranged from 81% to 92%. For early pregnancy, late pregnancy, and postpartum maternal plasma samples, the sequence reads across CpG sites were equal to average haploid coverage of 33-fold, 23-fold, and 19-fold per strand, respectively. Bisulfite conversion efficiency was >99.9% for all samples (Table 100).

In Table 100, the ambiguity rate (labeled "a") refers to the proportion of reads that simultaneously align to both the Watson and Crick chains of the reference human genome. The λ conversion rate refers to the proportion of unmethylated cytosine residues converted to "thymine" by bisulfite treatment of the internal control λDNA. H is generally equal to A, C, or T. "a" refers to reads that can be mapped to a specific locus but cannot be assigned to either the Watson or Crick chain. "b" refers to paired reads with consistent start and end coordinates. For "c", λDNA is added to each sample prior to bisulfite conversion. The λ conversion rate refers to the proportion of cytosine nucleotides converted to "thymine" after bisulfite conversion and is used as an indicator of successful bisulfite conversion. "d" refers to the number of cytosine nucleotides present in the reference human genome that retain a cytosine sequence after bisulfite conversion.

During bisulfite modification, unmethylated cytosine is converted to uracil and subsequently to thymine after PCR amplification, while methylated cytosine remains intact (M Frommer et al., 1992 Proc Natl Acad Sci USA; 89:1827-31). After sequencing and alignment, the methylation status of individual CpG sites can be inferred from the counts of methylated sequence reads "M" (methylated) and unmethylated sequence reads "U" (unmethylated) of cytosine residues in the CpG sequence. Using bisulfite sequencing data, the entire methylome of maternal blood, placenta, and maternal plasma is constructed. The average methylated CpG density (also known as methylation density MD) at a specific locus in maternal plasma can be calculated using the following equation:

Where M is the count of methylated reads at CpG sites within the locus and U is the count of unmethylated reads. If there is more than one CpG site within the locus, then M and U correspond to the counts across these sites.

B. Various technologies

As described above, methylation morphology analysis can be performed using massively parallel sequencing (MPS) of bisulfite-converted plasma DNA. MPS of bisulfite-converted plasma DNA can be performed randomly or using a shotgun approach. The sequencing depth can vary depending on the size of the relevant region.

In another embodiment, relevant regions in bisulfite-converted plasma DNA can be captured first using a solution-phase or solid-phase hybridization process, followed by MPS. Massive parallel sequencing can be performed using semiconductor-based sequencing systems such as the ilumina sequencing-by-synthesis platform, the SOLiD platform from Life Technologies, semiconductor-based sequencing systems such as the Ion Torrent or Ion Proton platforms from Life Technologies, or single-molecule sequencing systems such as the Helicos system or the Pacific Biosciences system, or nanopore-based sequencing systems. Nanopore-based sequencing includes the use of nanopores constructed from, for example, lipid bilayers and protein nanopores, as well as solid-state nanopores (e.g., graphene-based nanopores). Because the chosen single-molecule sequencing platform will allow for the direct detection of the methylation status of DNA molecules (including N6-methyladenine, 5-methylcytosine, and 5-hydroxymethylcytosine) without bisulfite conversion (BA Flusberg et al. 2010 Nat Methods; 7: 461-465; J Shim et al. 2013 Sci Rep; 3:1389.doi:10.1038/srep01389), the use of such a platform will allow for the analysis of the methylation status of DNA samples (e.g., plasma DNA) that have not undergone bisulfite conversion.

Besides sequencing, other techniques can be used. In one embodiment, methylation pattern analysis can be performed by methylation-specific PCR, or by PCR followed by digestion with a methylation-sensitive restriction enzyme, or by PCR followed by conjugation enzyme chain reaction. In other embodiments, the PCR is in the form of single-molecule or digital PCR (B Vogelstein et al., 1999 Proc Natl Acad Sci USA; 96: 9236-9241). In other embodiments, the PCR can be real-time PCR. In other embodiments, the PCR can be multiplex PCR.

II. Analysis of Methylation Group

Some embodiments can use whole-genome bisulfite sequencing to determine the methylation pattern of plasma DNA. Fetal methylation patterns can be determined by sequencing maternal plasma DNA samples, as described below. Therefore, fetal DNA molecules (and the fetal methylome) are non-invasively obtained during pregnancy, and changes are continuously monitored as pregnancy progresses. Because the sequencing data is comprehensive, the maternal plasma methylome can be studied at single-nucleotide resolution on a whole-genome scale.

Because the genomic coordinates of the sequencing reads are known, these data can be used to study the overall methylation level of the methylome or any relevant region in the genome and to make comparisons between different genetic elements. Furthermore, multiple sequence reads cover each CpG site or locus. Descriptions of some measures used to measure the methylome are currently provided.

A. Methylation of plasma DNA molecules

DNA molecules exist in human plasma in low concentrations and in fragmented forms, typically similar in length to a single nucleosome unit (YMD Lo et al. 2010 Sci Transl Med; 2:61ra91; and YW Zheng et al. 2012 Clin Chem; 58:549-558). Despite these limitations, whole-genome bisulfite sequencing pipelines are still able to analyze the methylation of plasma DNA molecules. In other embodiments, because the chosen single-molecule sequencing platform will allow direct inference of the methylation state of DNA molecules without bisulfite conversion (Fullaberg et al. 2010 Natural Methods; 7:461-465; Smoe et al. 2013 Scientific Reports; 3:1389. doi:10.1038/srep01389), using such a platform will allow sample DNA that has not undergone bisulfite conversion to be used to determine the methylation level of plasma DNA or to determine the plasma methylome. Such platforms can detect N6-methyladenine, 5-methylcytosine, and 5-hydroxymethylcytosine, which can provide improved results (e.g., increased sensitivity or specificity) corresponding to different forms of methylation associated with different biological functions. Such improved results can be applied to the detection or monitoring of specific conditions such as preeclampsia or certain types of cancer.

Bisulfite sequencing can also distinguish between different forms of methylation. In one embodiment, it may include an additional step that can differentiate between 5-methylcytosine and 5-hydroxymethylcytosine. One such method is oxobissulfite sequencing (oxBS-seq), which can elucidate the positions of 5-methylcytosine and 5-hydroxymethylcytosine at single-base resolution (Booth et al. 2012 Science; 336:934-937; Booth et al. 2013 Nature Protocols; 8:1841-1851). In bisulfite sequencing, both 5-methylcytosine and 5-hydroxymethylcytosine are read as cytosine and therefore cannot be distinguished. On the other hand, in oxBS-seq, 5-hydroxymethylcytosine is specifically oxidized to 5-formylcytosine by treatment with potassium perruthenate (KRuO4), followed by bisulfite conversion to convert the newly formed 5-formylcytosine to uracil, allowing 5-hydroxymethylcytosine to be distinguished from 5-methylcytosine. Therefore, 5-methylcytosine reads can be obtained from a single oxBS-seq operation, and the 5-hydroxymethylcytosine level can be inferred by comparison with bisulfite sequencing results. In another embodiment, 5-methylcytosine can be distinguished from 5-hydroxymethylcytosine using Tet-assisted bisulfite sequencing (TAB-seq) (M Yu et al. 2012 Nat Protoc; 7:2159-2170). TAB-seq can identify 5-hydroxymethylcytosine at single-base resolution and determine its abundance at each modification site. This method involves β-glucosyltransferase-mediated protection (glucosylation) of 5-hydroxymethylcytosine and recombinant mouse Tet1 (mTet1)-mediated oxidation of 5-methylcytosine to 5-carboxycytosine. Following subsequent bisulfite treatment and PCR amplification, both cytosine and 5-carboxycytosine (derived from 5-methylcytosine) are converted to thymine (T), while 5-hydroxymethylcytosine is read as C.

Figure 1B illustrates the methylation density within a 1Mb window of a sequencing sample according to an embodiment of the present invention. Figure 150 is a Circos diagram depicting the methylation density in maternal plasma and genomic DNA within a 1Mb window spanning the genome. From outside to inside: Chromosome G-banding can be arranged clockwise (centromeres shown in red) with pter-qter orientation, maternal blood (red), placenta (yellow), maternal plasma (green), shared readings in maternal plasma (blue), and fetal-specific readings in maternal plasma (purple). The overall CpG methylation levels (i.e., density levels) of maternal blood cells, placenta, and maternal plasma are shown in Table 100. Across the entire genome, the methylation level of maternal blood cells is generally higher than that of the placenta.

B. Comparison of bisulfite sequencing with other technologies

The placental methylome was studied using massively parallel bisulfite sequencing. Furthermore, the placental methylome was studied using an oligonucleotide array platform (Illumina) covering approximately 480,000 CpG sites in the human genome (Kurlis et al. 2012 Nature Genetics; 44:1236-1242; and Clark et al. 2012 PLoS One; 7:e50233). In one embodiment using bead-chip-based genotyping and methylation analysis, genotyping was performed using the Illumina HumanOmni2.5-8 genotyping array, according to the manufacturer's protocol. Genotyping was performed using the GenCall algorithm of Genome Studio Software (Illumina). The detection rate exceeded 99%. For microarray-based methylation analysis, following the manufacturer's recommendations for the Illumina Infinium Methylation Assay, genomic DNA (500-800 ng) was treated with sodium bisulfite using the Zymo EZ DNA Methylation Kit (Zymo Research, Orange, CA, USA).

According to the Infinium HD methylation analysis protocol, methylation analysis was performed on 4 μl of bisulfite-converted genomic DNA at 50 ng/μl. Hybridized bead arrays were scanned on an Ilumina iScan instrument. DNA methylation data were analyzed using the GenomeStudio (v2011.1) methylation module (v1.9.0) software, with normalization relative to internal controls and background subtraction. The methylation index of individual CpG sites is expressed as a β value (β), which is calculated using the ratio of fluorescence intensity between methylated and unmethylated alleles.

For CpG sites present on the array and sequenced to at least 10-fold coverage, compare the β value obtained by the array with the methylation index determined by sequencing the same site. The β value is the ratio of the intensity of the methylated probe to the total intensity of the methylated and unmethylated probes covering the same CpG site. The methylation index for each CpG site is the ratio of the methylated readings covering that CpG to the total readings.

Figures 2A-2C show scatter plots of β values determined by the Ilumina Infinim Human Methylation 450K Bead Array against methylation indices (measured by two platforms) determined by whole-genome bisulfite sequencing of corresponding CpG sites: (A) maternal hematopoietic cells, (B) chorionic villus samples, and (C) term placental tissue. The data from both platforms are highly consistent, and the Pearson correlation coefficients for maternal hematopoietic cells, CVS, and term placental tissue are 0.972, 0.939, and 0.954, respectively, with ^R² values of 0.945, 0.882, and 0.910.

The sequencing data were further compared with those reported by Chu et al., who used an oligonucleotide array covering approximately 27,000 CpG sites to study the methylation patterns of 12 pairs of CVS and maternal blood cell DNA samples (Chu et al. 2011 PLoS One; 6:e14723). Correlation analysis between the sequencing results of CVS and maternal blood cell DNA and each of the 12 pairs of samples in previous studies showed a mean Pearson coefficient (0.967) and ^R² (0.935) for maternal blood and a mean Pearson coefficient (0.943) and ^R² (0.888) for CVS. Among the CpG sites represented on both arrays, the data were highly correlated with published data. The non-CpG methylation rate of maternal blood cells, CVS, and placental tissue was <1% (Table 100). These results are consistent with current reports that a large amount of non-CpG methylation is primarily restricted in pluripotent cells (Lister et al. 2009 Nature; 462:315-322; Laurent et al. 2010 Genome Research; 20:320-331).

C. Comparison of plasma and blood methylomes in non-pregnant individuals

Figures 3A and 3B show bar charts of the percentage of methylated CpG sites in plasma and blood cells collected from adult men and non-pregnant adult women: (A) autosomes, (B) chromosome X. The figures illustrate the similarity between plasma and blood methylation groups in men and non-pregnant women. The overall proportion of methylated CpG sites in plasma samples from men and non-pregnant women was almost identical to that in the corresponding blood cell DNA (Table 100 and Figures 2A and 2B).

The correlation of methylation patterns in plasma and blood cell samples was then investigated in a locus-specific manner. Methylation density per 100kb region in the human genome was determined by identifying the proportion of unconverted cytosine at CpG sites relative to all CpG sites covered by sequence reads up to 100kb. Methylation density showed a high degree of agreement between plasma samples from both male and female subjects and their corresponding blood cell DNA.

Figures 4A and 4B show scatter plots of methylation density at corresponding loci in hematopoietic cell DNA and plasma DNA: (A) non-pregnant adult women, (B) adult men. The Pearson correlation coefficient and ^R² values for the non-pregnant female samples were 0.963 and 0.927, respectively, while those for the male samples were 0.953 and 0.908, respectively. These data are consistent with previous findings based on genotypic assessments of plasma DNA molecules from recipients of allogeneic hematopoietic ^stem cell transplantation, demonstrating that hematopoietic cells are the primary source of DNA in human plasma (Zheng et al. 2012 Clinical Chemistry; 58:549-558).

D. Methylation level across methylation groups

Subsequently, the methylation levels of maternal plasma DNA, maternal blood cells, and placental tissue DNA were studied to determine the methylation levels. These levels were determined for repeating regions, non-repeating regions, and the whole genome.

Figures 5A and 5B show bar charts of the percentage of methylated CpG sites in samples collected from pregnant women: (A) autosomes, (B) chromosome X. The total percentages of methylated CpG in maternal plasma samples from early and late pregnancy were 67.0% and 68.2%, respectively. Unlike results obtained from non-pregnant individuals, these percentages are lower than in early pregnancy maternal blood cell samples but higher than in CVS and term placental tissue samples (Table 100). Notably, the percentage of methylated CpG in postpartum maternal plasma samples was 73.1%, similar to the blood cell data (Table 100). These tendencies are observed in CpGs distributed across all autosomes and chromosome X, and across non-repetitive regions and multiple classes of repetitive elements in the human genome.

The study found that both repeating and non-repeating elements in the placenta were hypomethylated relative to maternal blood cells. This result is consistent with previous findings that the placenta is hypomethylated relative to other tissues, including peripheral blood cells.

CpG sites were sequenced with 71% to 72% methylation in DNA from blood cells of pregnant women, non-pregnant women, and adult men (Table 100, Figure 1). These data are comparable to the reported 68.4% methylation of CpG sites in blood monocytes, as reported by Li et al., 2010, PLOS Biology; 8:e1000533. Consistent with previous reports on the hypomethylation of placental tissue, 55% and 59% of CpG sites were methylated in CVS and term placental tissue, respectively (Table 100).

Figure 6 is a bar chart showing the methylation levels of different repeat classes in the human genome from maternal blood, placenta, and maternal plasma. Repeat classes are defined as defined by the UCSC genome browser. The data shown are from early pregnancy samples. Unlike earlier reports, which primarily observed hypomethylation of placental tissue in certain repeat classes within the genome (Novakovic et al. 2012 Placenta; 33:959-970), this chart shows that, relative to the genome of blood cells, the placenta actually exhibits hypomethylation in most repeat classes.

E. Similarity of methylation groups

The examples used the same platform to determine the methylome of placental tissue, blood cells, and plasma. Therefore, direct comparisons of the methylomes of those biological sample types are possible. The high similarity between the methylomes of blood cells and plasma from male and non-pregnant women, and between maternal blood cells and postpartum maternal plasma samples, further confirms that hematopoietic cells are the primary source of DNA in human plasma (Zheng et al. 2012 Clinical Chemistry; 58:549-558).

The similarity is evident based on the overall proportion of methylated CpGs in the genome and the high correlation between methylation densities at corresponding loci in blood cell DNA and plasma DNA. However, the overall proportion of methylated CpGs in maternal plasma samples from early and late pregnancy is lower than in maternal blood cell data or postpartum maternal plasma samples. The reduced methylation levels during pregnancy are due to the hypomethylated nature of fetal DNA molecules present in maternal plasma.

The reversal of methylation patterns in postpartum maternal plasma samples to become more similar to maternal blood cells indicates that fetal DNA molecules have been removed from maternal circulation. Calculations of fetal DNA concentration based on fetal SNP markers indeed show a change in concentration from 33.9% prenatally to only 4.5% in postpartum samples.

F. Other Applications

The examples have successfully assembled the DNA methylome using MPS analysis of plasma DNA. The ability to determine the placental or fetal methylome from maternal plasma provides a non-invasive method for measuring, detecting, and monitoring abnormal methylation patterns associated with pregnancy-related conditions such as preeclampsia, intrauterine growth restriction, and preterm birth. For example, the detection of disease-specific abnormal methylation markers allows for the screening, diagnosis, and monitoring of such pregnancy-related conditions. Measurement of maternal plasma methylation levels allows for the screening, diagnosis, and monitoring of such pregnancy-related conditions. In addition to its direct application in research on pregnancy-related conditions, the method can also be applied to other medical fields where plasma DNA analysis is of interest. For example, the methylome of cancer can be determined from the plasma DNA of cancer patients. As described in this article, cancer methylome analysis from plasma may be a technique that works in conjunction with cancer genomic analysis from plasma (KCA et al. 2013 Clin Chem; 59:211-224 and RJ Leary et al. 2012 Sci Transl Med; 4:162ra154).

For example, the determination of methylation levels in plasma samples can be used for cancer screening. When the methylation levels in a plasma sample show abnormal levels compared to healthy controls, cancer can be suspected. The type of cancer or the origin of the cancer tissue can then be further confirmed and evaluated by measuring the plasma methylation patterns at different loci or by using plasma genomic analysis to detect tumor-related copy number abnormalities, chromosomal translocations, and single nucleotide variants. In fact, in one embodiment of the invention, plasma cancer methylome and genomic morphology analysis can be performed simultaneously. Alternatively, radiological and imaging studies (e.g., computed tomography, magnetic resonance imaging, positron emission tomography) or endoscopic examinations (e.g., upper gastrointestinal endoscopy or colonoscopy) can be used based on plasma methylation level analysis to further investigate individuals suspected of having cancer.

For cancer screening or detection, the determination of methylation levels in plasma (or other biological) samples can be combined with other modalities used for cancer screening or detection, such as: prostate-specific antigen measurements (e.g. for prostate cancer), carcinoembryonic antigen (e.g. for colorectal cancer, gastric cancer, pancreatic cancer, lung cancer, breast cancer, medullary thyroid carcinoma), alpha-fetoprotein (e.g. for liver cancer or germ cell tumors), CA125 (e.g. for ovarian and breast cancer), and CA19-9 (e.g. for pancreatic cancer).

Additionally, sequencing can be performed on other tissues to obtain the cellular methylome. For example, liver tissue can be analyzed to determine specific methylation patterns for the liver, which can be used to identify liver lesions. Other tissues that can be analyzed include brain cells, bone, lungs, heart, muscles, and kidneys. The methylation patterns of various tissues may change over time due to development, aging, disease processes (such as inflammation or cirrhosis, or autoimmune processes (such as in systemic lupus erythematosus)), or treatment (such as treatment with demethylating agents such as 5-azacytidine and 5-azadeoxycytidine). The dynamic nature of DNA methylation makes such analyses potentially valuable for monitoring physiological and pathological processes. For example, if an individual's plasma methylome is detected to have changed compared to baseline values obtained when they are healthy, then disease progression in organs can subsequently be detected because organs release DNA into the plasma.

Furthermore, the methylome of transplanted organs can be determined from the plasma DNA of organ transplant recipients. As described in this invention, transplant methylome analysis from plasma may be a technique that works in conjunction with transplant genomic analysis from plasma (YW Zheng et al., 2012; YMD Lo et al., 1998 Lancet; 351:1329-1330; and TM Snyder et al., 2011 Proc Natl Acad Sci USA; 108:6229-6234). Because plasma DNA is generally considered a marker of cell death, an increase in plasma levels of DNA released from a transplanted organ can be used as a marker of increased cell death in that organ, such as in cases of rejection or other pathological processes (e.g., infection or abscess). In cases where anti-rejection therapy is successfully initiated, the plasma levels of DNA expected to be released from the transplanted organ are reduced.

III. Using SNPs to determine fetal or tumor methylation profiles

As mentioned above, for non-pregnant individuals, the plasma methylome corresponds to the blood methylome. However, for pregnant women, these methylomes differ. Fetal DNA molecules circulate in maternal plasma against a background of predominantly maternal DNA (YMD Lo et al., 1998, *American Journal of Human Genetics*; 62:768-775). Therefore, for pregnant women, the plasma methylome is essentially a complex of the placental methylome and the blood methylome. Consequently, the placental methylome can be extracted from plasma.

In one embodiment, single nucleotide polymorphism (SNP) differences between mother and fetus are used to identify fetal DNA molecules in maternal plasma. The goal is to identify SNP loci where the mother is homozygous but the fetus is heterozygous; fetal-specific alleles can be used to determine which DNA fragment originates from the fetus. Genomic DNA from maternal blood cells is analyzed using the Ilumina HumanOmni2.5-8 SNP genotyping array. On the other hand, for SNP loci where the mother is heterozygous and the fetus is homozygous, maternal-specific SNP alleles can then be used to determine which plasma DNA fragment originates from the mother. The methylation level of such DNA fragments will reflect the methylation level of the relevant genomic region in the mother.

A. Correlation between fetal-specific methylation readings and placental methylome

Sequencing results from biological samples identify loci with two distinct alleles, where one allele (B) is significantly less abundant than the other (A). Readings covering the B allele are considered fetal-specific (fetal-specific readings). If the mother is determined to be homozygous for A and the fetus to be heterozygous for A/B, then readings covering the A allele are shared by both mother and fetus (shared readings).

In an analysis of a pregnancy case used to illustrate several concepts in this invention, the pregnant mother was found to be homozygous at 1,945,516 loci on autosomes. Maternal plasma DNA sequencing reads covering these SNPs were examined. Reads with non-maternal alleles were detected at 107,750 loci, and these loci were considered informative loci. At each informative SNP, the allele not from the mother is called the fetal-specific allele, while the other is called the shared allele.

The percentage concentration of fetal/tumor DNA in maternal plasma (also known as fetal DNA percentage) can be determined. In one embodiment, the percentage concentration f of fetal DNA in maternal plasma is determined by the following equation:

Where p is the number of sequencing reads for fetal-specific alleles, and q is the number of sequencing reads for shared alleles between mother and fetus (Lou et al. 2010 Science Translational Medicine; 2:61ra91). The proportion of fetal DNA in maternal plasma samples from early pregnancy, late pregnancy, and postpartum was found to be 14.4%, 33.9%, and 4.5%, respectively. The proportion of fetal DNA was also calculated using the number of reads aligned to chromosome Y. Based on chromosome Y data, the results were 14.2%, 34.9%, and 3.7% in maternal plasma samples from early pregnancy, late pregnancy, and postpartum, respectively.

By analyzing fetal-specific or shared sequence reads separately, the examples demonstrated that circulating fetal DNA molecules are significantly less methylated than background DNA molecules. Comparison of methylation density at corresponding loci in fetal-specific maternal plasma reads with placental tissue data revealed a high correlation between the two in early and late pregnancy. These data confirm at the genomic level that the placenta is the primary source of fetal-derived DNA molecules in maternal plasma and represent a significant step forward from previous evidence based on information derived from individual selected loci.

Methylation density in each 1Mb region of the genome was determined using fetal-specific or shared reads covering CpG sites near informative SNPs. Fetal and non-fetal-specific methylomes assembled from maternal plasma sequence reads can be visualized, for example, using Circos diagrams (MKrzywinski et al. 2009 Genome Res; 19:1639-1645). Methylation density in each 1Mb region was also determined from maternal blood cell and placental tissue samples.

Figure 7A shows a Circos diagram 700 for an early pregnancy sample. Figure 7B shows a Circos diagram 750 for a late pregnancy sample. Figures 700 and 750 show the methylation density per 1 Mb interval. The chromosome G-banding diagram (outermost ring) is pter-qter oriented clockwise (centromeres are shown in red). The second outermost trajectory shows the number of CpG sites corresponding to the 1 Mb region. The scale of the red bars shown is up to 20,000 sites per 1 Mb interval. The methylation density corresponding to the 1 Mb region is shown in the other rings according to the color arrangement of the central position.

For early pregnancy samples (Figure 7A), from the inside out, the different rings correspond to: chorionic villus sampling, fetal-specific readings in maternal plasma, maternal-specific readings in maternal plasma, total fetal and non-fetal readings in maternal plasma, and maternal blood cells. For late pregnancy samples (Figure 7B), the different rings correspond to: term placental tissue, fetal-specific readings in maternal plasma, maternal-specific readings in maternal plasma, total fetal and non-fetal readings in maternal plasma, postpartum maternal plasma, and maternal blood cells (from early pregnancy blood samples). It can be observed that for both early and late pregnancy plasma samples, the fetal methylation group is more hypomethylated than the non-fetal-specific methylation group.

The overall methylation pattern of the fetal methylome was more similar to that of CVS or placental tissue samples. Conversely, the DNA methylation pattern of shared reads in plasma, which was primarily maternal DNA, was more similar to that of maternal blood cells. A systematic, locus-by-locus comparison of methylation density between maternal plasma DNA reads and maternal or fetal tissue was then performed. We identified sites with methylation density at CpG sites on the same sequence reads as informative SNPs and which were covered by at least five maternal plasma DNA sequence reads.

Figures 8A-8D show a comparison of methylation density of genomic tissue DNA relative to maternal plasma DNA at CpG sites surrounding informative single nucleotide polymorphisms. Figure 8A shows the methylation density of fetal-specific readings in early pregnancy maternal plasma samples relative to CVS readings. It can be seen that the fetal-specific values correspond very well to the CVS values.

Figure 8B shows the methylation density of fetal-specific readings in maternal plasma samples during late pregnancy, relative to readings in full-term placental tissue. Again, the methylation density group corresponds very well to the readings in full-term placental tissue, indicating that fetal methylation patterns can be obtained by analyzing readings with fetal-specific alleles.

Figure 8C shows the methylation density of shared readings in early pregnancy maternal plasma samples relative to maternal blood cell readings. It is assumed that most shared readings originate from the mother, and the two sets of values correspond very well. Figure 8D shows the methylation density of shared readings in late pregnancy maternal plasma samples relative to maternal blood cell readings.

For fetal-specific readings in maternal plasma, the Spearman correlation coefficient between early pregnancy maternal plasma and CVS was 0.705 (P < 2.2*e⁻¹⁶); and the Spearman correlation coefficient between late pregnancy maternal plasma and term placental tissue was 0.796 (P < 2.2*e⁻¹⁶) (Figures 8A and 8B). Shared readings in maternal plasma were compared similarly with maternal blood cell data. The Pearson correlation coefficient for early pregnancy plasma samples was 0.653 (P < 2.2*e⁻¹⁶) and the Pearson correlation coefficient for late pregnancy plasma samples was 0.638 (P < 2.2*e⁻¹⁶) (Figures 8C and 8D).

B. Fetal methylome

In one embodiment, to collect a fetal methylome from maternal plasma, sequence reads spanning at least one informative fetal SNP site and containing at least one CpG site within the same read are sorted. Reads exhibiting fetal-specific alleles are included in the collection of fetal methylomes. Reads exhibiting shared alleles, i.e., non-fetal-specific alleles, are included in the collection of non-fetal-specific methylomes consisting primarily of maternally derived DNA molecules.

For maternal plasma samples from early pregnancy, fetal-specific readings covered 218,010 CpG loci on autosomes. The corresponding figures for late pregnancy and postpartum maternal plasma samples were 263,611 and 74,020, respectively. On average, shared readings covered those CpG loci an average of 33.3, 21.7, and 26.3 times, respectively. For early pregnancy, late pregnancy, and postpartum maternal plasma samples, fetal-specific readings covered those CpG loci an average of 3.0, 4.4, and 1.8 times, respectively.

Fetal DNA represents a minority of CpG sites in maternal plasma, and therefore the coverage of those CpG sites by fetal-specific readings is proportional to the percentage of fetal DNA in the sample. For early pregnancy maternal plasma samples, the total percentage of methylated CpGs in fetal readings was 47.0%, compared to 68.1% in shared readings. For late pregnancy maternal plasma samples, the percentage of methylated CpGs in fetal readings was 53.3%, compared to 68.8% in shared readings. These data demonstrate that fetal-specific readings in maternal plasma are less methylated than shared readings in maternal plasma.

C. Method

The above techniques can also be used to determine tumor methylation patterns. Methods for determining fetal and tumor methylation patterns are now described.

Figure 9 is a flowchart illustrating a method 900 for determining a first methylation pattern from a biological sample of an organism according to an embodiment of the present invention. Method 900 can construct an epigenetic map of the fetus from the methylation patterns of maternal plasma. The biological sample comprises cell-free DNA containing a mixture of cell-free DNA derived from a first tissue and a second tissue. As an example, the first tissue may be derived from a fetus, a tumor, or a transplanted organ.

In box 910, multiple DNA molecules from a biological sample are analyzed. The analysis of DNA molecules may include determining the location of the DNA molecule within the organism's genome, determining the genotype of the DNA molecule, and determining whether the DNA molecule is methylated at one or more sites.

In one embodiment, DNA molecules are analyzed using sequence reads, where sequencing is methylation-identifiable sequencing. Therefore, the sequence reads include the methylation state of DNA molecules from a biological sample. The methylation state may include whether a specific cytosine residue is 5-methylcytosine or 5-hydroxymethylcytosine. Sequence reads can be obtained from various sequencing technologies, PCR technologies, arrays, and other techniques suitable for identifying fragment sequences. The methylation state of the sites in the sequence reads can be obtained as described herein.

At box 920, multiple first loci are identified where the first genome of a first tissue is heterozygous for the corresponding first allele and the corresponding second allele, and the second genome of a second tissue is homozygous for the corresponding first allele. For example, fetal-specific reads can be identified at multiple first loci. Or tumor-specific reads can be identified at multiple first loci. Tissue-specific reads can be identified from sequencing reads where the percentage of second allele sequence reads is within a specific range, such as approximately 3%–25%, indicating that a minority of DNA fragments originate from heterozygous genomes at the loci, while the majority originate from homozygous genomes at the loci.

At box 930, DNA molecules located at one or more sites at each first locus are analyzed. The number of DNA molecules methylated at the site and corresponding to the corresponding second allele at the locus is determined. Each locus may contain more than one site. For example, an SNP may indicate that a fragment is fetal-specific, and the fragment may have multiple sites that define the methylation status. The number of methylation reads at each site can be determined, and the total number of methylation reads at the locus can be measured.

A locus can be defined by a specific number of sites, a specific group of sites, or a specific size of the region surrounding a variant containing tissue-specific alleles. A locus may have only one site. Sites may have specific characteristics, such as being CpG sites. The determination of the number of unmethylated reads is equivalent and is included within the determination of methylation status.

At box 940, for each first locus, the methylation density is calculated based on the number of DNA molecules methylated at one or more sites at the locus and corresponding to the corresponding second allele at the locus. For example, the methylation density can be determined for the CpG site corresponding to the locus.

At box 950, the first methylation pattern of the first tissue is derived from the methylation density of the first locus. The first methylation pattern may correspond to a specific site, such as a CpG site. The methylation pattern may be targeted at all loci with fetal-specific alleles or only some of those loci.

IV. Differences in plasma and blood methylation groups

The above has demonstrated the correlation between fetal-specific readings from plasma and the placental methylome. Because the maternal component of the maternal plasma methylome is primarily contributed by blood cells, the difference between the plasma methylome and the blood methylome can be used to determine the placental methylome for all loci, not just the location of fetal-specific alleles. The difference between the plasma methylome and the blood methylome can also be used to determine the methylome of tumors.

A. Method

Figure 10 is a flowchart illustrating a method 1000 for determining a first methylation pattern from a biological sample of an organism according to an embodiment of the present invention. The biological sample (e.g., plasma) comprises a mixture of cell-free DNA derived from a first tissue and a second tissue. The first methylation pattern corresponds to the methylation pattern of the first tissue (e.g., fetal tissue or tumor tissue). Method 1200 can infer regions of differential methylation from maternal plasma.

At box 1010, a biological sample is received. The biological sample can be simply received on a machine (e.g., a sequencer). The biological sample can be in the form of collection from an organism or in a processed form, such as plasma extracted from a blood sample.

At box 1020, the second methylation pattern corresponding to the DNA of the second tissue is obtained. The second methylation pattern can be read from memory, as it may have been previously determined. The second methylation pattern can be determined from the second tissue, such as different samples containing only or primarily cells of the second tissue. The second methylation pattern can correspond to the cellular methylation pattern and be obtained from cellular DNA. As another example, the second pattern can be determined from plasma samples collected before pregnancy or before the onset of cancer, because the plasma methylation profile of non-pregnant individuals without cancer is very similar to the methylation profile of blood cells.

The second methylation pattern can provide the methylation density of each of multiple loci in an organism's genome. The methylation density at a particular locus corresponds to the proportion of methylated DNA in the second tissue. In one embodiment, the methylation density is the CpG methylation density, where the CpG site associated with the locus is used to determine the methylation density. If a locus has only one site, then the methylation density can be equal to the methylation index. The methylation density also corresponds to the unmethylated density because the two values are complementary.

In one embodiment, the second methylation pattern is obtained by sequencing regions of methylation-recognizable DNA from cellular DNA from a biological sample. An example of sequencing that recognizes methylation includes treating DNA with sodium bisulfite followed by DNA sequencing. In another example, sequencing that recognizes methylation can be performed without sodium bisulfite using a single-molecule sequencing platform that would allow direct detection of the methylation status of DNA molecules (including N6-methyladenine, 5-methylcytosine, and 5-hydroxymethylcytosine) without bisulfite conversion (Fullerberg et al. 2010 Nature Methods; 7:461-465; Smoe et al. 2013 Science Reports; 3:1389. doi: 10.1038/srep01389); or via... Immunoprecipitation of permethylated cytosine (e.g., using antibodies against methylcytosine or using methylated DNA-binding proteins or peptides (LG Acevedo et al. 2011 Epigenomics; 3: 93-101)) followed by sequencing; or using methylation-sensitive restriction enzymes followed by sequencing. In another embodiment, non-sequencing techniques such as array, digital PCR, and mass spectrometry are used.

In another embodiment, the second methylation density of the second tissue can be obtained in advance from a control sample of the individual or from other individuals. The methylation density from another individual can serve as a reference methylation pattern with a reference methylation density. The reference methylation density can be determined from multiple samples, wherein the average level (or other statistical value) of different methylation densities at the locus can be used as the reference methylation density at said locus.

At box 1030, the free methylation pattern is determined from the free DNA of the mixture. The free methylation pattern provides the methylation density at each of the multiple loci. The free methylation pattern can be determined by receiving sequence reads from the sequencing of the free DNA, where methylation information is obtained from the sequence reads. The free methylation pattern can be determined in the same manner as the cellular methylome.

At box 1040, the percentage of cell-free DNA from a first tissue in the biological sample is determined. In one embodiment, the first tissue is fetal tissue, and the corresponding DNA is fetal DNA. In another embodiment, the first tissue is tumor tissue, and the corresponding DNA is tumor DNA. The percentage can be determined in various ways, such as using fetal-specific alleles or tumor-specific alleles. Copy number can also be used to determine the percentage, as described, for example, in U.S. Patent Application 13/801,748, filed March 13, 2013, entitled "Mutational Analysis of Plasma DNA for Cancer Detection" (incorporated by reference).

At box 1050, multiple loci are identified for determining the first methylation group. These loci may correspond to each locus used to determine the free methylation pattern and the second methylation pattern. Therefore, multiple loci may correspond. It is possible that even more loci can be used to determine the free methylation pattern and the second methylation pattern.

In some embodiments, maternal blood cells can be used, for example, to identify loci that are hypermethylated or hypomethylated in the second methylation pattern. To identify hypermethylated loci in maternal blood cells, CpG sites with a methylation index ≥ X% (e.g., 80%) can be scanned from one end of the chromosome. The next CpG site in a downstream region (e.g., within 200 bp downstream) can then be searched. If the immediately downstream CpG site also has a methylation index ≥ X% (or other specified amount), then the first and second CpG sites can be merged into one group. Merging can continue until no other CpG sites exist in a downstream region, or the methylation index of the immediately downstream CpG site is < X%. If the region of the merged CpG sites contains at least five adjacent hypermethylated CpG sites, then the region can be reported as hypermethylated in maternal blood cells. Similar analyses can be performed to search for hypomethylated loci in maternal blood cells for CpG sites with a methylation index ≤ 20%. The methylation density of the second methylation type of the shortlisted loci can be calculated and used, for example, to infer the first methylation type of the corresponding loci (e.g., placental tissue methylation density) from maternal plasma bisulfite sequencing data.

At box 1060, the first methylation type of the first tissue is determined by calculating a difference parameter, including the difference between the methylation density of the second methylation type and the methylation density of the free methylation type, for each of the multiple loci. The difference is measured as a percentage.

In one embodiment, the first methylation density (D) of a locus in a first (e.g., placental) tissue is inferred using the following equation:

Where mbc represents the methylation density of the second methylation pattern at the locus (e.g., a shortlisted locus identified from maternal blood cell bisulfite sequencing data); mp represents the methylation density of the corresponding locus in maternal plasma bisulfite sequencing data; f represents the percentage of cell-free DNA from the first tissue (e.g., the percentage concentration of fetal DNA); and CN represents the copy number at the locus (e.g., higher for amplification or lower for deletion relative to normal). If there is no amplification or deletion in the first tissue, then CN can be one. For triploidy (or chromosomal repetitive regions in tumors or fetuses), CN will be 1.5 (because it increases from 2 copies to 3 copies), and for haploidy it will be 0.5. Higher amplification can be increased in increments of 0.5. In this example, D can correspond to the differential parameter.

At box 1070, the first methylation density is transformed to obtain the corrected first methylation density of the first tissue. The transformation may result in a fixed difference between the difference parameter and the actual methylation pattern of the first tissue. For example, the values may differ by a fixed constant or slope. The transformation can be linear or non-linear.

In one embodiment, the distribution of the inferred value D was found to be lower than the actual methylation level in placental tissue. For example, the inferred value can be obtained using a linear transformation of data from CpG islands, which are genomic regions with a relatively high proportion of CpG sites. The genomic locations of the CpG islands used in this study were obtained from the UCSC Genome Browser database (NCBI build 36/hg18) (Fujita et al. 2011 Nucleic Acids Res; 39:D876-882). For example, a CpG island can be defined as a genomic region with a GC content ≥50%, a genome length >200bp, and an observed/expected CpG number ratio >0.6 (M Gardiner-Garden et al., 1987, J Mol Biol; 196: 261-282).

In one implementation, to extrapolate the linear transformation equation, CpG islands in the sequencing sample with at least four CpG sites and an average read depth ≥5 for each CpG site can be included. After determining the linear relationship between the methylation density of CpG islands in CVS or full-term placenta and the inferred value D, the predicted value is determined using the following equation:

Early pregnancy prediction value = D × 1.6 + 0.2

Predictive value for late pregnancy = D × 1.2 + 0.05

B. Fetal Case

As mentioned above, Method 1000 can be used to infer the methylation profile of the placenta from maternal plasma. Circulating DNA in plasma primarily originates from hematopoietic cells. An unknown proportion of cell-free DNA remains, contributed by other internal organs. Furthermore, placental-derived cell-free DNA accounts for approximately 5-40% of total DNA in maternal plasma, with an average of approximately 15%. Therefore, it can be assumed that the methylation level in maternal plasma corresponds to background methylation plus the contribution of the placenta during pregnancy, as described above.

The maternal plasma methylation level (MP) can be determined using the following equation:

MP = BKG × (1-f) + PLN × f

BKG represents the background DNA methylation level in plasma derived from blood cells and internal organs, PLN represents the methylation level of the placenta, and f represents the percentage concentration of fetal DNA in maternal plasma.

In one embodiment, the methylation level of the placenta can be theoretically derived as follows:

Equations (1) and (2) are equal when CN equals one, D equals PLN, and BKG equals mbc. In another embodiment, the percentage concentration of fetal DNA can be assumed or set to a specified value, for example, as part of the assumption of a minimum f.

The methylation level of maternal blood was obtained to represent the background methylation of maternal plasma. In addition to loci with high or low methylation in maternal blood cells, inference methods were further explored by focusing on clinically relevant delimited regions, such as CpG islands in the human genome.

The average methylation density of a total of 27,458 CpG islands (NCBI Build36/hg18) on autosomes and chromosome X was derived from sequencing data from maternal plasma and placenta. Only CpG islands covering ≥10 CpG sites and having an average read depth ≥5 for each covered site were selected from all analyzed samples, including placenta, maternal blood, and maternal plasma. As a result, 26,698 CpG islands (97.2%) remained valid, and their methylation levels were inferred using plasma methylation data and the percentage concentration of fetal DNA derived from the equation above.

It is noted that the distribution of the inferred PLN values is lower than the actual methylation level of CpG islands in placental tissue. Therefore, in one embodiment, the inferred PLN value, or simple inferred value (D), is used as an arbitrary unit to assess the methylation level of CpG islands in the placenta. After transformation, the inferred values become linear and their distribution becomes more similar to the actual dataset. The transformed inferred values are named methylation prediction values (MPV) and are subsequently used to predict the methylation level of loci in the placenta.

In this example, CpG islands were categorized into three classes based on their methylation density in the placenta: low (≤0.4), medium (>0.4–<0.8), and high (≥0.8). Using inference equations, the MPV of CpG islands within the same group was calculated, and they were subsequently categorized into the three classes using the same threshold corresponding to the stated values. By comparing the actual and inferred datasets, it was found that 75.1% of the CpG islands shortlisted based on MPV values correctly matched the same category in the tissue data. Approximately 22% of CpG islands were assigned to groups with a level 1 difference (high vs. medium, or medium vs. low), and less than 3% were completely misclassified (high vs. low) (Figure 12A). Overall classification performance was also determined: 86.1%, 31.4%, and 68.8% of CpG islands with methylation densities ≤0.4, >0.4–<0.8, and ≥0.8 in the placenta, respectively, were correctly inferred as “low,” “medium,” and “high” (Figure 12B).

Figures 11A and 11B illustrate the performance of an algorithm for predicting percentage concentrations of fetal DNA using maternal plasma data according to an embodiment of the present invention. Figure 11A, which is Figure 1100, shows the accuracy of CpG island classification using MPV-corrected classification (inferred class accurately matches the actual dataset), level 1 discrepancy (inferred class differs from the actual dataset by 1 level), and misclassification (inferred class is opposite to the actual dataset). Figure 11B, which is Figure 1150, shows the proportion of CpGs correctly classified in each inferred class.

Assuming low maternal background methylation in the corresponding genomic regions, the presence of highly methylated placental-derived DNA in circulation will increase total plasma methylation levels to a degree dependent on the percentage concentration of fetal DNA. Significant changes can be observed when the released fetal DNA is fully methylated. Conversely, when maternal background methylation is high, changes in plasma methylation levels will be more significant if hypomethylated fetal DNA is released. Therefore, inference schemes may be more practical when inferring methylation levels from loci known to differ between the maternal background and placenta, particularly from markers of hypermethylation and hypomethylation in the placenta.

Figure 12A is Table 1200, illustrating details of methylation prediction for 15 selected loci according to an embodiment of the invention. To validate the technique, 15 previously studied differentially methylated loci were selected. The methylation levels of the selected regions were inferred and compared with those of the 15 previously studied differentially methylated loci (RWK Chiu et al. 2007 American Journal of Pathology; 170:941-950; S.S.C. Chim et al. 2008 Clinical Chemistry; 54:500-511). ; Zhan et al. 2005 Proc Natl Acad Sci U.S.A.; 102:14753-14758; Cui et al. 2010 PloS One; 5:e15069.

Figure 12B is Figure 1250, showing the inferred methylation levels of 15 selected loci in the placenta and their corresponding methylation levels. The inferred methylation levels are: low (≤0.4), medium (>0.4-<0.8), and high (≥0.8). Table 1200 and Figure 1300 show that the methylation levels in the placenta can be appropriately inferred, with a few exceptions: RASSF1A, CGI009, CGI137, and VAPA. Of these four markers, only CGI009 shows a significant misclassification compared to the actual dataset. The others are minor misclassifications.

In Table 1200, “1” refers to the inferred value (D) calculated using the following equation: where f is the percentage concentration of fetal DNA. “2” refers to the methylation prediction value (MPV) of the inferred value, using a linear transformation of the following equation: MPV = D × 1.6 + 0.25. “3” indicates the threshold classification corresponding to the inferred value: low, ≤0.4; medium, >0.4-<0.8; high, ≥0.8. “4” indicates the threshold classification corresponding to the actual placental dataset: low, ≤0.4; medium, >0.4-<0.8; high, ≥0.8. “5” indicates that the placental status refers to the placental methylation status relative to maternal blood cells.

C. Calculate the percentage concentration of fetal DNA

In one embodiment, the percentage of fetal DNA from the first tissue can be derived from the Y chromosome of a male fetus. The proportion of chromosome Y (%chrY) sequence in a maternal plasma sample is a complex of the number of chromosome Y readings derived from a male fetus and the number of misaligned chromosome Y readings in the mother (female) (Qiu et al. 2011 BMJ; 342: c7401). Therefore, the relationship between %chrY in the sample and the percentage concentration (f) of fetal DNA can be given as follows:

%chrY _male refers to the proportion of readings that align with chromosome Y in plasma samples containing 100% male DNA; and %chrY _female refers to the proportion of readings that align with chromosome Y in plasma samples containing 100% female DNA.

%chrY can be determined by mismatch-free alignment of readings to chromosome Y with samples from women carrying male fetuses, for example, where the readings are from bisulfite-converted samples. %chrY _male values can be obtained from bisulfite sequencing of plasma samples from two adult males. %chrY _female values can be obtained from bisulfite sequencing of plasma samples from two non-pregnant adult women.

In other embodiments, the percentage of fetal DNA can be determined by fetal-specific alleles on autosomes. As another example, epigenetic markers can be used to determine the percentage of fetal DNA. Other methods for determining the percentage of fetal DNA may also be used.

D. Methods for determining copy number using methylation

The placental genome is more hypomethylated than the maternal genome. As discussed above, the methylation of maternal plasma depends on the percentage concentration of placental-derived fetal DNA in the maternal plasma. Therefore, by analyzing the methylation density of chromosomal regions, differences in the contribution of fetal tissue to maternal plasma can be detected. For example, in pregnant women with trisomy (e.g., trisomy 21, 18, or 13), the fetus will contribute an additional amount of DNA from the trisomy to the maternal plasma compared to a disomy. In this case, the plasma methylation density of the trisomy (or any chromosomal region with amplification) will be lower than that of the disomy. The degree of difference can be predicted mathematically by considering the percentage concentration of fetal DNA in the plasma sample. The higher the percentage concentration of fetal DNA in the plasma sample, the greater the difference in methylation density between trisomy and disomy chromosomes. For regions with deletions, the methylation density will be even higher.

One example of a missing chromosome is Turner syndrome, in which the female fetus will have only one copy of chromosome X. In this case, the methylation density of chromosome X in the plasma DNA of a pregnant woman carrying a fetus with Turner syndrome will be higher than that of the same pregnant woman carrying a female fetus with a normal number of chromosome X chromosomes. In one embodiment of this strategy, the presence or absence of chromosome Y sequence in maternal plasma can be analyzed first (e.g., using MPS or PCR-based techniques). If chromosome Y sequence is present, the fetus can be classified as male and the following analysis is unnecessary. On the other hand, if chromosome Y sequence is absent in maternal plasma, the fetus can be classified as female. In this case, the methylation density of chromosome X in maternal plasma can then be analyzed. A higher than normal chromosome X methylation density will indicate a high risk of Turner syndrome in the fetus. This method can also be applied to other sex chromosome aneuploidies. For example, for a fetus with XYY chromosomes, the methylation density of the Y chromosome in maternal plasma will be lower than that of a normal XY fetus with similar levels of fetal DNA in maternal plasma. As another example, in fetuses with Klinefelter syndrome (XXY), the Y chromosome sequence is present in the maternal plasma, but the methylation density of chromosome X in the maternal plasma will be lower than that of normal XY fetuses with similar fetal DNA levels in the maternal plasma.

Based on the previous discussion, the plasma methylation density (MP _euploidy ) of disomy chromosomes can be calculated as: MP _euploidy = BKG × (1-f) + PLN × f, where BKG is the background DNA methylation level in plasma derived from blood cells and viscera, PLN is the methylation level of the placenta, and f is the percentage concentration of fetal DNA in maternal plasma.

The plasma methylation density (MP _aneuploidy ) of trisomic chromosomes can be calculated as: MP _aneuploidy = BKG × (1-f) + PLN × f × 1.5, where 1.5 corresponds to the copy number CN, and each additional chromosome increases the density by 50%. The difference between trisomic and disomy chromosomes (MP _Diff ) will be...

MP _Diff = PLN × f × 0.5.

In one embodiment, comparing the methylation density of a potentially aneuploid chromosome (or chromosomal region) with the total methylation density of one or more other assumed euploid chromosomes or genomes can be used to effectively normalize fetal DNA concentration in a plasma sample. The comparison can be used to obtain a normalized methylation density by calculating a parameter (e.g., involving a ratio or difference) between the methylation densities of the two regions. The comparison can remove dependence on the resulting methylation level (e.g., a parameter determined from the two methylation densities).

If the methylation density of a potentially aneuploid chromosome is not normalized relative to the methylation density of one or more other chromosomes or other parameters reflecting the percentage concentration of fetal DNA, then the percentage concentration will be the primary factor influencing plasma methylation density. For example, the plasma methylation density of chromosome 21 in a pregnant woman carrying a fetus with trisomy 21 and a fetal DNA percentage concentration of 10% will be the same as that in a pregnant woman carrying an euploid fetus and a fetal DNA percentage concentration of 15%, whereas normalized methylation densities will show differences.

In another embodiment, the methylation density of potentially aneuploid chromosomes can be normalized relative to the percentage concentration of fetal DNA. For example, the following equation can be applied to normalize methylation density: where MP _normalized is the methylation density normalized to the percentage concentration of fetal DNA in plasma, MP _unnormalized is the measured methylation density, BKG is the background methylation density from maternal blood cells or tissue, PLN is the methylation density in placental tissue, and f is the percentage concentration of fetal DNA. The methylation densities of BKG and PLN can be based on reference values previously established from maternal blood cells and placental tissue obtained from healthy pregnant women. Different genetic and epigenetic methods can be used, for example, to determine the percentage concentration of fetal DNA in a plasma sample by measuring the percentage of sequence readings from chromosome Y on DNA that has not undergone bisulfite conversion using massively parallel sequencing or PCR.

In one implementation, the normalized methylation density of aneuploid chromosomes may be compared to a reference population consisting of pregnant women carrying euploid fetuses. The mean and standard deviation (SD) of the normalized methylation density of the reference population can be determined. Subsequently, the normalized methylation density of the test cases can be expressed as a z-score, indicating the number of SDs relative to the reference population mean: where MP is the _normalized methylation density of the test cases, the mean is the average of the normalized methylation densities of the reference cases, and the SD is the standard deviation of the normalized methylation density of the reference cases. For example, a threshold of z-score < -3 can be used to classify whether chromosomes are significantly hypomethylated, and thus determine the aneuploid state of the sample.

In another embodiment, the MP _Diff can be used as a normalized methylation density. In such embodiments, the PLN can be inferred, for example, using method 1000. In some implementations, a reference methylation density (which can be normalized using f) can be determined from the methylation level of the euploid region. For example, the average value can be determined from one or more chromosomal regions of the same sample. The threshold can be measured using f, or simply set to a level sufficient to present a minimum concentration.

Therefore, comparisons between the methylation level of a region and a threshold can be made in various ways. Comparisons can include standardization (e.g., as described above), which can be applied equally to either the methylation level or the threshold, depending on how the value is defined. Thus, it can be determined in several ways whether a region's determined methylation level differs statistically from a reference level (determined from the same sample or other samples).

The above analysis can be applied to the analysis of chromosomal regions, which may include whole chromosomes or portions of chromosomes, including adjacent or separate subregions of chromosomes. In one embodiment, a possible aneuploid chromosome can be divided into a large number of regions. The fractional fetal DNA concentrations may have the same or different sizes. The methylation density of each region can be normalized relative to a percentage concentration of the sample or relative to the methylation density of one or more assumed euploid chromosomes or the total methylation density of the genome. The normalized methylation density of each region can then be compared with a reference population to determine whether it is significantly hypomethylated. The percentage of significantly hypomethylated regions can then be determined. For example, thresholds of 5%, 10%, 15%, 20%, or 30% of significantly hypomethylated regions can be used to classify the aneuploid state of the case.

When testing for amplification or deletion, the methylation density can be compared to a reference methylation density that may be specific to the particular region being tested. Each region can have a different reference methylation density because methylation can vary with region, particularly depending on the size of the region (e.g., the smaller the region, the more varied it is).

As mentioned above, one or more pregnant women, each carrying an euploid fetus, can be used to define the normal range of methylation density in the relevant region or the difference in methylation density between two chromosomal regions. The normal range of PLN can also be determined (e.g., by direct measurement or inference as in method 1000). In other embodiments, the ratio between two methylation densities, such as the ratio between two methylation densities of a possible aneuploid chromosome and an euploid chromosome, can be used for analysis rather than their difference. This methylation analysis method can be combined with sequence read counting methods (RWK Chiu et al. 2008 Proc Natl Acad Sci USA; 105:20458-20463) and methods involving plasma DNA size analysis (US Patent 2011/0276277) to identify or confirm aneuploidy. Sequence read counting methods used in conjunction with methylation analysis can employ random sequencing (Qiu et al., 2008, Proceedings of the National Academy of Sciences; 105:20458-20463; Bianchi et al., 2012, Journal of Obstetrics and Gynecology 119:890-901) or targeted sequencing (AB Sparks et al., 2012, Journal of Obstetrics and Gynecology 206:319.e1-9). et al. 2012 Am J Obstet Gynecol 206:319.e1-9); Zimmermann et al. 2012 Prenatal Diagnosis 32:1233-1241; Liao et al. 2012 PloS One; 7:e38154.

The use of BKG can account for background variations between samples. For example, one woman may have different BKG methylation levels than another, but in such cases, the difference between BKG and PLN can be used across samples. Thresholds can differ for different chromosomal regions, for example, when the methylation density of one region of the genome differs from that of another region of the genome.

This method can be extended to detect any chromosomal abnormalities in the fetal genome, including deletions and amplifications. Furthermore, the resolution of this analysis can be adjusted to desired levels; for example, the genome can be divided into 10Mb, 5Mb, 2Mb, 1Mb, 500kb, and 100kb regions. Therefore, this technique can also be used to detect subchromosomal duplications or deletions. Consequently, this technique will allow for non-invasive prenatal fetal molecular karyotype acquisition. When used in this manner, this technique can be combined with non-invasive prenatal testing methods based on molecular counting (ASrinivasan et al. 2013 American Journal of Human Genetics; 92:167-176; SCY Yu et al. 2013 PLOS One 8:e60968). In other embodiments, the region sizes do not need to be uniform. For example, the size of the regions can be adjusted so that each region contains the same number of dinucleotides. In this case, the actual size of the regions will be different.

The equation can be rewritten as MP _Diff = (BKG - PLN) × f × 0.5 × CN to accommodate different types of chromosomal abnormalities. Here, CN represents the number of copy number changes in the affected region. For a chromosome with one copy increase, CN equals 1; for a chromosome with two copies increase, CN equals 2; and for the loss of one of two homologous chromosomes (e.g., used to detect fetal Turner syndrome, where a female fetus loses one of its X chromosomes, resulting in an XO karyotype), CN equals -1. The equation does not need to be changed when the region size changes. However, sensitivity and specificity may decrease when using smaller region sizes because smaller regions contain fewer CpG dinucleotides (or other combinations of nucleotides that demonstrate differential methylation between fetal and maternal DNA), leading to increased random variation in the measurement of methylation density. In one embodiment, the number of readings required can be determined by analyzing the coefficient of variation of methylation density and the required level of sensitivity.

To illustrate the feasibility of this method, plasma samples from nine pregnant women were analyzed. Five of the pregnant women carried euploid fetuses, and the other four each carried fetuses with trisomy 21 (T21). Three of the five euploid pregnancies were randomly selected to form a reference population. The remaining two euploid pregnancies (Eu1 and Eu2) and four T21 cases (T21-1, T21-2, T21-3, and T21-4) were analyzed using this method to test for possible T21 statistic. Plasma DNA was bisulfite-converted and sequenced using the ilumina HiSeq2000 platform. In one embodiment, the methylation density of individual chromosomes was calculated. The difference in methylation density between chromosome 21 and the mean of the other 21 pairs of autosomes was then determined to obtain normalized methylation density (Table 1). The mean and SD of the reference population were used to calculate the z-scores for the six test cases.

Table 1: Using a threshold of <-3 for z-scores, samples were classified as T21, and all euploid and T21 cases were correctly classified.

In another embodiment, the genome is divided into 1Mb regions and the methylation density of each 1Mb region is determined. The methylation density of all regions on a hypoploid chromosome can be normalized to the median methylation density of all regions on a hypothetical euploid chromosome. In one implementation, for each region, the difference between the methylation density and the median value of the euploid region can be calculated. The z-scores of these values can be calculated using the mean and SD values of a reference population. The percentage of regions exhibiting low methylation can be determined (Table 2) and compared to a threshold percentage.

Table 2: Using 5% as a threshold for regions that are significantly more hypomethylated on chromosome 21, all cases were correctly classified for the T21 state.

Methods for detecting fetal chromosomal or subchromosomal abnormalities based on this DNA methylation can be combined with methods based on, for example, sequencing (Qiu et al., 2008, Proceedings of the National Academy of Sciences; 105:20458-20463), digital PCR (Luo et al., 2007, Proceedings of the National Academy of Sciences; 104:13116-13121), or DNA molecule size determination (US Patent Publication 2011/0276277) for molecule counting. Such combinations (e.g., DNA methylation plus molecule counting, or DNA methylation plus size determination, or DNA methylation plus molecule counting plus size determination) will have synergistic effects that will be advantageous in a clinical setting, such as improving sensitivity and/or specificity. For example, the number of DNA molecules that would need to be analyzed, for example, by sequencing, can be reduced without adversely affecting diagnostic accuracy. This feature will allow such tests to be performed more economically. As another example, for a given number of DNA molecules analyzed, the combined methods will allow for the detection of fetal chromosomal or subchromosomal abnormalities at a lower percentage concentration of fetal DNA.

Figure 13 is a flowchart of method 1300 for detecting chromosomal abnormalities from a biological sample of an organism. The biological sample includes cell-free DNA comprising a mixture of cell-free DNA derived from a first tissue and a second tissue. The first tissue may be derived from a fetus or a tumor, and the second tissue may be derived from a pregnant woman or a patient.

At box 1310, multiple DNA molecules from a biological sample are analyzed. The analysis of DNA molecules may include determining the location of DNA molecules within the organism's genome and determining whether the DNA molecules are methylated at one or more sites. The analysis can be performed by receiving sequence reads from sequencing that identifies methylation, thus the analysis can be performed solely on data previously obtained from the DNA. In other embodiments, the analysis may include actual sequencing or other active steps to obtain data.

Location determination may include mapping DNA molecules (e.g., by sequence readings) to corresponding parts of the human genome, such as specific regions. In one implementation, if a reading does not align to the relevant region, then the reading can be ignored.

At box 1320, for each of a plurality of sites, the corresponding number of DNA molecules methylated at that site is determined. In one embodiment, the site is a CpG site, and may be simply a number of CpG sites, selected using one or more criteria mentioned herein. Once normalized using the total number of DNA molecules analyzed at a particular site, such as the total number of sequence reads, the number of methylated DNA is equivalent to determining the number of unmethylated DNA.

At box 1330, the first methylation level of the first chromosomal region is calculated based on the corresponding number of site-methylated DNA molecules within the first chromosomal region. The first chromosomal region can have any size, such as those described above. The methylation level can take into account the total number of DNA molecules compared to the first chromosomal region, for example, as part of a normalization procedure.

The first chromosomal region can be of any size (e.g., the entire chromosome) and can be composed of separate subregions, i.e., subregions separated from each other. The methylation level of each subregion can be determined and combined to, for example, an average or median value to determine the methylation level of the first chromosomal region.

At box 1340, the first methylation level is compared to a threshold. The threshold can be a reference methylation level or a value related to the reference methylation level (e.g., a specified distance from the normal level). The threshold can be determined from other pregnant women carrying a fetus who have no chromosomal abnormalities in the first chromosome region, from samples from individuals without cancer, or from loci (i.e., regions of disomy) in organisms known not to be associated with aneuploidy.

In one embodiment, the threshold can be defined as the difference between (BKG - PLN) × f × 0.5 × CN and a reference methylation level, where BKG is the female background (or the mean or median from other individuals), f is the percentage concentration of cell-free DNA derived from the first tissue, and CN is the copy number being tested. CN is an instance of a correction factor corresponding to a type of aberration (deletion or duplication). An initial threshold of CN = 1 can be used to test all amplifications, and subsequent thresholds can be used to determine the extent of amplification. The threshold can be based on the percentage concentration of cell-free DNA derived from the first tissue to determine the expected level of methylation at the locus, for example, in the absence of copy number aberrations.

At box 1350, the classification of abnormalities in the first chromosomal region is determined based on comparison. Different levels of statistical significance can indicate an increased risk of a fetus with chromosomal abnormalities. In various embodiments, the chromosomal abnormality may be trisomy 21, trisomy 18, trisomy 13, Turner syndrome, or Klinefelter syndrome. Other examples are subchromosomal deletions, subchromosomal duplications, or DiGeorge syndrome.

V. Determination of markers

As mentioned above, the methylation of certain parts of the fetal genome differs from that of the maternal genome. These differences are likely common in pregnant women. Regions with different methylations can be used to identify DNA fragments originating from the fetus.

A. Methods for determining DMR from placental and maternal tissues

The placenta possesses tissue-specific methylation markers. Fetal-specific DNA methylation markers have been developed based on loci with differential methylation between placental tissue and maternal blood cells for maternal plasma detection and non-invasive prenatal diagnostic applications (Zhan et al. 2008 Clinical Chemistry; 54:500-511; Papageorgiou et al. 2009 American Journal of Pathology; 174:1609-1618; and Zhu et al. 2011 PLoS One; 6:e14723). Examples of identifying such differentially methylated regions (DMRs) on a genome-wide basis are provided.

Figure 14 is a flowchart of method 1400, which, according to embodiments of the invention, identifies methylation markers by comparing placental methylation patterns with maternal methylation patterns (e.g., as determined from blood cells). Method 1400 can also be used to determine tumor markers by comparing tumor methylation patterns with methylation patterns corresponding to healthy tissue.

At box 1410, the placental methylome and blood methylome are obtained. The placental methylome can be determined from placental samples, such as CVS or full-term placenta. It should be understood that the methylome may include the methylation density of only a portion of the genome.

At box 1420, the region is identified, comprising a specified number of sites (e.g., 5 CpG sites) and for which a sufficient number of reads have been obtained. In one embodiment, identification begins at one end of each chromosome to locate the first 500 bp region containing at least five valid CpG sites. If a CpG site is covered by at least five sequence reads, then the site is considered valid.

At box 1430, the placental methylation index and blood methylation index for each site are calculated. For example, the methylation index is calculated separately for all eligible CpG sites within each 500 bp region.

At box 1440, methylation indices are compared between maternal blood cell and placental samples to determine whether the indices of the groups are different from each other. For example, the Mann-Whitney test is used to compare methylation indices between maternal blood cells and CVS or term placenta. For example, a p-value ≤0.01 is considered statistically significant, although other values can be used, where lower values will reduce the false positive area.

In one embodiment, if the number of valid CpG sites is less than five or the Man-Whitney test is not significant, the 500bp region is shifted downstream by 100bp. The region continues to shift downstream until the Man-Whitney test becomes significant for the 500bp region. The next 500bp region is then considered. If the next region is found to be statistically significant by the Man-Whitney test, it is added to the current region, but the combined adjacent regions are no greater than 1,000bp.

At box 1450, adjacent regions that are statistically significantly different (e.g., by the Mann-Whitney test) can be merged. Note that the difference is between the methylation indices of the two samples. In one embodiment, adjacent regions are merged if they are within a specified distance (e.g., 1,000 bp) from each other and if they exhibit similar methylation patterns. In one implementation, the similarity of methylation patterns between adjacent regions can be defined using any of the following: (1) showing the same tendency in placental tissue relative to maternal blood, e.g., both regions are more methylated in placental tissue than in blood cells; (2) the difference in methylation density between adjacent regions in placental tissue is less than 10%; and (3) the difference in methylation density between adjacent regions in maternal blood cells is less than 10%.

At box 1460, the methylation density of blood methylome from maternal blood cell DNA and placental samples (e.g., CVS or full-term placental tissue) in the region is calculated. Methylation density can be determined as described herein.

At box 1470, candidate DMRs are identified based on a statistically significant difference between the total placental methylation density and the total blood methylation density across all sites in the region. In one embodiment, all valid CpG sites within the merged region undergo a ^χ² test. For all valid CpG sites within the merged region, the ^χ² test assesses whether there is a statistically significant difference between maternal blood cells and placental tissue when the number of methylated cytosines is presented as the ratio of methylated to unmethylated cytosines. In one implementation, a p-value ≤ 0.01 is considered statistically significant for the ^χ² test. Merged segments demonstrating significance through the ^χ² test are considered candidate DMRs.

At box 1480, loci with maternal blood cell DNA methylation density exceeding a high threshold or falling below a low threshold are identified. In one embodiment, loci with maternal blood cell DNA methylation density ≤20% or ≥80% are identified. In other embodiments, bodily fluids other than maternal blood may be used, including (but not limited to) saliva, uterine or cervical lavage fluid from the female reproductive tract, tears, sweat, saliva, and urine.

A key to successfully developing fetal-specific DNA methylation markers from maternal plasma may be ensuring that the methylation state of maternal blood cells is as high as possible or unmethylated. This reduces (e.g., minimizes) the likelihood that maternal DNA molecules interfere with the analysis of placental-derived fetal DNA molecules exhibiting the opposite methylation pattern. Therefore, in one embodiment, candidate DMRs are selected through further filtering. Candidate hypomethylated loci are loci exhibiting a methylation density ≤20% in maternal blood cells and a methylation density at least 20% higher in placental tissue. Candidate hypermethylated loci are loci exhibiting a methylation density ≥80% in maternal blood cells and a methylation density at least 20% lower in placental tissue. Other percentages may be used.

At box 1490, DMR is then identified in a subset of loci where placental methylation density differs significantly from blood methylation density by comparing the difference to a threshold. In one embodiment, the threshold is 20%, so the methylation density differs from the methylation density from maternal blood cells by at least 20%. Therefore, the difference between placental methylation density and blood methylation density can be calculated for each identified locus. The difference can be a simple subtraction. In other embodiments, correction factors and other functions can be used to determine the difference (e.g., the difference can be the result of a function applied to a simple subtraction).

In one implementation, using this method, 11,729 hypermethylated and 239,747 hypomethylated loci were identified from placental samples from early pregnancy. The top 100 hypermethylated loci are listed in Table S2A of the Annex. The top 100 hypomethylated loci are listed in Table S2B of the Annex. Tables S2A and S2B list the chromosome, start and end positions, region size, methylation density in maternal blood, methylation density in placental samples, p-values (all very small), and methylation differences. The positions correspond to the reference genome hg18, which can be found at hgdownload.soe.ucsc.edu/goldenPath/hg18/chromosomes.

11,920 hypermethylated and 204,768 hypomethylated loci were identified from placental samples in late pregnancy. The top 100 hypermethylated loci in late pregnancy are listed in Table S2C, and the top 100 hypomethylated loci are listed in Table S2D. Thirty-three loci previously reported to show differential methylation between maternal blood cells and early pregnancy placental tissue were used to validate a list of early pregnancy candidates. Using our algorithm, 79% of the 33 loci were identified as DMRs.

Figure 15A is Table 1500, demonstrating the performance of the DMR identification algorithm using early pregnancy data on 33 previously reported early pregnancy markers. In this table, "a" indicates loci 1 to 15 previously described in (RWK Chiu et al. 2007 American Journal of Pathology; 170:941-950 and Zhan et al. 2008 Clinical Chemistry; 54:500-511); loci 16 to 23 previously described in (Yuan, Paper 2007). The data are derived from the above-mentioned published literature. The data are from the sequencing data generated in this study, but based on genomic coordinates provided by the initial study. (KC Yuen, thesis 2007, The Chinese University of Hong Kong, Hong Kong); and loci 24 to 33 were previously described in (EA Papageorgiou et al. 2009 Am J Pathol; 174:1609-1618). “b” indicates that these data are derived from the above-mentioned published literature. “c” indicates that the methylation density and its differences in maternal blood cells and chorionic villus samples were observed from sequencing data generated in this study, but based on genomic coordinates provided by the initial study. “d” indicates that the data regarding the locus was identified using the embodiment of Method 1400 on bisulfite sequencing data, without reference to the published literature cited above by Qiu et al. (2007), Zhan et al. (2008), Yuan (2007), and Papayeo et al. (2009). The span of the locus includes previously reported genomic regions, but generally spans larger regions. “e” indicates that, based on the requirement of observing a difference >0.20 between the methylation density of the corresponding genomic coordinates of the DMR in maternal hematopoietic cells and chorionic villi samples, the candidate DMR is classified as a true positive (TP) or a false negative (FN).

Figure 15B is Table 1550, demonstrating the performance of the DMR identification algorithm using late pregnancy data and compared with placental samples obtained at delivery. “a” indicates the use of the same list of 33 loci as described in Figure 17A. “b” indicates that because these 33 loci were previously identified from early pregnancy samples, they may not be applicable to late pregnancy data. Therefore, in this study, bisulfite sequencing data corresponding to term placental tissue were re-examined based on the genomic coordinates provided in the initial study. A difference in methylation density >0.20 between maternal hematopoietic cells and term placental tissue was used to determine whether these loci were indeed true DMRs in late pregnancy. “c” indicates that the data for these loci were identified using Method 1400 on bisulfite sequencing data without reference to previously cited publications by Qiu et al. (2007), Zhan et al. (2008), Yuan (2007), and Papayeo et al. (2009). The span of the loci includes previously reported genomic regions but generally spans larger areas. The "d" designation indicates that candidate DMRs containing loci identified as differentially methylated in late pregnancy are classified as true positive (TP) or false negative (FN) based on a requirement that the difference in methylation density between the corresponding genomic coordinates of the DMR observed in maternal blood cells and full-term placental tissue is >0.20. Loci not identified as differentially methylated in late pregnancy, those not present in the DMR list, or those containing these loci but showing a methylation difference <0.20, are considered true negative (TN) DMRs.

B. DMR from maternal plasma sequencing data

It should be possible to directly identify placental tissue DMR from maternal plasma DNA bisulfite sequencing data, assuming the percentage concentration of fetal DNA in the sample is also known. This is likely because the placenta is the primary source of fetal DNA in maternal plasma (Chim et al., 2005 Proc Natl Acad SciUSA 102, 14753-14758), and this study demonstrated that the methylation status of fetal-specific DNA in maternal plasma is correlated with the placental methylome.

Therefore, aspects of method 1400 can be achieved by using plasma methylome to determine the inferred placental methylome instead of using placental samples. Thus, methods 1000 and 1400 can be combined to determine the DMR. Method 1000 can be used to determine the predicted value of placental methylation morphology and used in method 1400. For this analysis, the example also focuses on loci in maternal blood cells with methylation ≤20% or ≥80%.

In one implementation, to infer loci in placental tissue that are hypermethylated relative to maternal blood cells, loci in maternal blood cells with methylation ≤20% and methylation ≥60% according to predicted values, and a difference of at least 50% between the blood cell methylation density and the predicted value, are sorted and displayed. To infer loci in placental tissue that are hypomethylated relative to maternal blood cells, loci in maternal blood cells with methylation ≥80% and methylation ≤40% according to predicted values, and a difference of at least 50% between the blood cell methylation density and the predicted value, are sorted and displayed.

Figure 16 is Table 1600, showing the number of loci predicted as hypermethylated or hypomethylated based on direct analysis of maternal plasma bisulfite sequencing data. "N/A" indicates not applicable. "a" indicates that the search for hypermethylated loci begins with a list of loci showing methylation density <20% in maternal blood cells. "b" indicates that the search for hypomethylated loci begins with a list of loci showing methylation density >80% in maternal blood cells. "c" indicates that bisulfite sequencing data from chorionic villus samples were used to validate early pregnancy maternal plasma data, and full-term placental tissue was used to validate late pregnancy maternal plasma data.

As shown in Table 1600, most of the non-invasively inferred loci exhibited the expected methylation patterns in tissue and overlapped with the DMRs found from tissue data and presented in the preceding sections. The appendix lists the DMRs identified from plasma. Table S3A lists the top 100 loci inferred as hypermethylated from early pregnancy maternal plasma bisulfite sequencing data. Table S3B lists the top 100 loci inferred as hypomethylated from early pregnancy maternal plasma bisulfite sequencing data. Table S3C lists the top 100 loci inferred as hypermethylated from late pregnancy maternal plasma bisulfite sequencing data. Table S3D lists the top 100 loci inferred as hypomethylated from late pregnancy maternal plasma bisulfite sequencing data.

C. Pregnancy changes in placental and fetal methylation groups

The total proportion of methylated CpGs in the CVS was 55%, compared to 59% in the term placenta (Table 100, Figure 1). More hypomethylated DMRs could be identified in the CVS than in the term placenta, while the number of hypermethylated DMRs was similar in both tissues. Therefore, it is evident that the CVS is more hypomethylated than the term placenta. This pregnancy-related methylation trend was also evident in maternal plasma data. The proportion of methylated CpGs in fetal-specific readings was 47.0% in early pregnancy maternal plasma, but 53.3% in late pregnancy maternal plasma. The number of validated hypermethylated loci was similar in early (1,457 loci) and late (1,279 loci) maternal plasma samples, but the early pregnancy (21,812 loci) sample had substantially more hypomethylated loci than the late pregnancy (12,677 loci) sample (Table 1600, Figure 16).

D. The purpose of markers

Differentially methylated markers, or DMRs, are applicable in several ways. The presence of such markers in maternal plasma indicates and confirms the presence of fetal or placental DNA. This confirmation can be used as quality control for non-invasive prenatal testing. DMRs can serve as universal fetal DNA markers in maternal plasma and are superior to markers that rely on genotypic differences between mother and fetus, such as polymorphism-based markers or Y-chromosome-based markers. DMRs are universal fetal markers applicable to all pregnant women. Polymorphism-based markers are only applicable to a subset of pregnant women in which the fetus inherits the marker from its father and the mother does not possess the marker in her genome. Furthermore, the concentration of fetal DNA in maternal plasma samples can be measured by quantifying DNA molecules derived from those DMRs. By knowing the expected DMR morphology for normal pregnant women, pregnancy-related complications, particularly those involving placental tissue alterations, can be detected by observing deviations in maternal plasma DMR or methylation morphology from the expected morphology for normal pregnant women. Pregnancy-related complications involving placental tissue alterations include (but are not limited to) fetal chromosomal aneuploidy. Examples include trisomy 21, preeclampsia, intrauterine growth retardation, and preterm birth.

E. Kits using markers

Examples may provide compositions and kits for practicing the methods described herein, as well as other applicable methods. The kits can be used to analyze fetal DNA, such as cell-free fetal DNA, in maternal plasma. In one embodiment, the kit may include at least one oligonucleotide suitable for specific hybridization with one or more loci identified herein. The kit may also include at least one oligonucleotide suitable for specific hybridization with one or more reference loci. In one embodiment, a marker of placental hypermethylation is measured. The test locus may be methylated DNA in maternal plasma, and the reference locus may be methylated DNA in maternal plasma. Similar kits can be formulated for analyzing tumor DNA in plasma.

In some cases, the kit may include at least two oligonucleotide primers that can be used to amplify at least a portion of a target locus (e.g., the loci listed in the appendix) and a reference locus. Instead of primers, the kit may also include labeled probes for detecting DNA fragments corresponding to the target and reference loci. In various embodiments, one or more oligonucleotides in the kit correspond to the loci listed in the appendix. Typically, the kit also includes an instruction manual guiding the user to analyze test samples and assess physiological or pathological conditions in the tested individual.

In various embodiments, kits are provided for analyzing fetal DNA in biological samples containing a mixture of fetal DNA and DNA from a pregnant female individual. The kit may contain one or more oligonucleotides for specific hybridization with at least a portion of the genomic regions listed in Tables S2A, S2B, S2C, S2D, S3A, S3B, S3C, and S3D. Therefore, any number of oligonucleotides, ranging from those spanning multiple tables to those from only one table, can be used. The oligonucleotides may act as primers and may be organized into primer pairs, where one pair corresponds to a specific region from a table.

VI. Relationship between size and methylation density

Plasma DNA molecules are known to exist in circulation as short molecules, with most being approximately 160 bp in length (Luo et al. 2010 Science Translational Medicine; 2:61ra91; Zheng et al. 2012 Clinical Chemistry; 58:549-558). Interestingly, the data revealed a relationship between the methylation state and size of plasma DNA molecules. Therefore, plasma DNA fragment length is associated with DNA methylation levels. The characteristic size morphology of plasma DNA molecules suggests that most are likely associated with mononuclear bodies, which may originate from enzymatic degradation during apoptosis.

Circulating DNA is essentially fragmented. Specifically, circulating fetal DNA in maternal plasma samples is shorter than maternally derived DNA (KCA Chan et al. 2004 Clin Chem; 50: 88-92). Because paired-end alignment allows for size analysis of bisulfite-treated DNA, it is possible to directly assess whether there is any correlation between the size of plasma DNA molecules and their corresponding methylation levels. This was explored in maternal plasma and control plasma samples from non-pregnant adult women.

In this study, paired-end sequencing (which includes sequencing the entire molecule) was used to analyze each sample. The length of the sequenced DNA molecule was determined by aligning the paired-end sequences of each DNA molecule to a reference human genome and recording the genomic coordinates of the very end of the sequencing reads. Plasma DNA molecules are naturally fragmented into small molecules, and sequencing libraries of plasma DNA are typically prepared without any fragmentation steps. Therefore, the length inferred from sequencing represents the size of the initial plasma DNA molecule.

Previous studies have determined the size morphology of fetal and maternal DNA molecules in maternal plasma (Lou et al. 2010 Science Translational Medicine; 2:61ra91). These studies showed that plasma DNA molecules are similar in size to mononuclear bodies, and that fetal DNA molecules are shorter than maternal DNA molecules. In this study, the relationship between the methylation state and size of plasma DNA molecules was determined.

A. Result

Figure 17A shows curve 1700, illustrating the size distribution of DNA in maternal plasma, non-pregnant female control plasma, placenta, and peripheral blood. For both maternal and non-pregnant female control plasma samples, the two bisulfite-treated plasma samples exhibited the same characteristic size distribution previously reported (Lo et al. 2010 Science Translational Medicine; 2:61ra91), with the largest total sequence lengths being 166–167 bp and 10 bp period DNA molecules being shorter than 143 bp.

Figure 17B shows the size distribution and methylation patterns of maternal plasma, adult female control plasma, placental tissue, and adult female control blood. The average methylation density was calculated for DNA molecules of the same size containing at least one CpG site. The relationship between DNA molecule size and methylation density was then plotted. Specifically, for sequencing reads covering at least one CpG site, the average methylation density of each fragment ranging in length from 50 bp to at most 180 bp was determined. Interestingly, methylation density increased with plasma DNA size, peaking at approximately 166–167 bp. However, this pattern was not observed in placental and control blood DNA samples fragmented using an ultrasound generator system.

Figure 18 shows graphs of methylation density and size of plasma DNA molecules. Figure 18A is curve 1800 for maternal plasma in early pregnancy. Figure 18B is curve 1850 for maternal plasma in late pregnancy. Data for all sequencing reads covering at least one CpG site are represented by the blue curve 1805. Data containing reads of fetal-specific SNP alleles are represented by the red curve 1810. Data containing reads of maternal-specific SNP alleles are represented by the green curve 1815.

Readings containing fetal-specific SNP alleles are considered to originate from fetal DNA molecules. Readings containing maternal-specific SNP alleles are considered to originate from maternal DNA molecules. Generally, DNA molecules with high methylation density are longer. This tendency is present in fetal and maternal DNA molecules in early and late pregnancy. The overall size of fetal DNA molecules is shorter than previously reported for maternal DNA molecules.

Figure 19A shows curve 1900 for methylation density and size of sequencing reads from adult non-pregnant women. Plasma DNA samples from adult non-pregnant women also show the same relationship between DNA molecule size and methylation state. On the other hand, genomic DNA samples were fragmented by a sonication step prior to MPS analysis. As shown in curve 1900, data from blood cell and placental tissue samples did not show the same trend. Because cell fragmentation is artificial, size is not expected to be related to density. Since naturally fragmented DNA molecules in plasma are size-dependent, it can be assumed that lower methylation density makes molecules more likely to break into smaller fragments.

Figure 19B, curve 1950, illustrates the size distribution and methylation patterns of fetal-specific and maternal-specific DNA molecules in maternal plasma. Fetal-specific and maternal-specific plasma DNA molecules also show the same correlation between fragment size and methylation level. Fragment lengths of cell-free DNA from both placental origin and maternal circulation increase with methylation level. Furthermore, the distributions of their methylation states do not overlap, indicating that this phenomenon is independent of the initial fragment length of the circulating DNA molecules.

B. Method

Therefore, size distribution can be used to assess the total methylation percentage of a plasma sample. This methylation measurement can then be tracked by continuous measurements of the size distribution of plasma DNA, according to the relationships shown in Figures 18A and 18B, during pregnancy, cancer surveillance, or treatment. Methylation measurements can also be used to look for increased or decreased DNA release from relevant organs or tissues. For example, specific DNA methylation markers specific to a particular organ (e.g., the liver) can be identified and their concentrations in plasma measured. Because DNA is released into the plasma upon cell death, an increase in levels may indicate increased cell death or damage in that particular organ or tissue. A decrease in levels in a particular organ may indicate that treatment to combat damage or pathological processes in that organ is under control.

Figure 20 is a flowchart of method 2000, which is used to assess the methylation level of DNA in a biological sample of an organism according to embodiments of the present invention. The methylation level can be assessed against a specific region of the genome or the entire genome. If a specific region is desired, then DNA fragments from only that specific region can be used.

At box 2010, the amount of DNA fragments corresponding to various sizes is measured. For each of the multiple sizes, the amount of multiple DNA fragments in the biological sample corresponding to said size can be measured. For example, the number of DNA fragments with a length of 140 bases can be measured. The amounts can be stored in the form of a histogram. In one embodiment, the size of multiple nucleic acids in the biological sample is measured, which can be done individually (e.g., by single-molecule sequencing of the whole molecule or only the ends of the molecule) or in groups (e.g., by electrophoresis). The size can correspond to a range. Therefore, the amount can be for DNA fragments with sizes within a specific range. When performing paired-end sequencing, the methylation level of said region can be determined using DNA fragments aligned to a specific region (e.g., determined by paired sequence readings).

At box 2020, a first value for the first parameter is calculated based on the amount of DNA fragments at multiple sizes. In one respect, the first parameter provides a statistical measurement (e.g., a histogram) of the size morphology of DNA fragments in a biological sample. The parameter may be referred to as a size parameter because it is determined by the sizes of multiple DNA fragments.

The first parameter can take various forms. One parameter is the percentage of DNA fragments of a specific size or size range relative to all DNA fragments or relative to DNA fragments of another size or size range. Such a parameter is the number of DNA fragments at a specific size divided by the total number of fragments, which can be obtained from a histogram (any data structure that provides an absolute or relative count of fragments at a specific size). As another example, the parameter can be the number of fragments at a specific size or within a specific range divided by the number of fragments at another size or size range. Division can serve as a form of standardization to account for the different numbers of DNA fragments analyzed for different samples. Standardization can be achieved by analyzing the same number of DNA fragments for each sample, which effectively provides the same result as dividing by the total number of fragments analyzed. Additional examples of parameter and size analysis can be found in U.S. Patent Application 13/789,553, which is incorporated by reference for all purposes.

At box 2030, a first size value is compared with a reference size value. The reference size value can be calculated from a DNA fragment of a reference sample. To determine the reference size value, the methylation pattern and the value of the first size parameter can be calculated and quantified for the reference sample. Therefore, when the first size value is compared with the reference size value, the methylation level can be determined.

At box 2040, the methylation level is evaluated based on comparison. In one embodiment, a first value of a first parameter can be determined to be higher or lower than a reference size value, thereby determining whether the methylation level of the sample of the present invention is higher or lower than the reference size value. In another embodiment, the comparison is implemented by inputting the first value into a calibration function. The calibration function can effectively compare the first value with calibration values (a set of reference size values) by means of points on the discrimination curve corresponding to the first value. The subsequently evaluated methylation level is provided as the output value of the calibration function.

Therefore, size parameters can be calibrated to methylation levels. For example, methylation levels can be measured and correlated with specific size parameters of the sample. Data points from various samples can then be fitted to a calibration function. In one implementation, different calibration functions can be used for different subsets of DNA. Therefore, based on prior knowledge of the relationship between methylation and size for specific subsets of DNA, several different forms of calibration may exist. For example, the calibration of fetal and maternal DNA might be different.

As shown above, the placenta and maternal blood have lower methylation, resulting in smaller fetal DNA due to lower methylation. Therefore, methylation density can be assessed using the average size (or other statistical values) of sample fragments. Because fragment size can be measured using paired-end sequencing, rather than the potentially more technically complex methylation-identifying sequencing, this method could be cost-effective in clinical applications. This method can be used to monitor methylation changes associated with pregnancy progression or pregnancy-related conditions such as preeclampsia, preterm birth, and fetal disorders (e.g., those caused by chromosomal or genetic abnormalities or intrauterine growth restriction).

In another embodiment, this method can be used to detect and monitor cancer. For example, with successful cancer treatment, the methylation patterns in plasma or another bodily fluid, as measured using this size-based method, will change towards the methylation patterns of a cancer-free healthy individual. Conversely, in the case of cancer progression, the methylation patterns in plasma or another bodily fluid will diverge from the methylation patterns of a cancer-free healthy individual.

In summary, hypomethylated molecules in plasma are shorter than hypermethylated molecules. The same tendency is observed in both fetal and maternal DNA molecules. Since DNA methylation is known to affect nucleosome filling, the data suggest that hypomethylated DNA molecules may be less densely wrapped around histones and therefore more sensitive to enzymatic degradation. On the other hand, the data presented in Figures 18A and 18B also show that although fetal DNA has much lower methylation than the maternal readings, the size distributions of fetal and maternal DNA are not entirely separated. In Figure 19B, it can be seen that even for the same size category, the methylation levels of fetal and maternal specific readings differ from each other. This observation suggests that the hypomethylation state of fetal DNA is not the only factor explaining its shorter length relative to maternal DNA.

VII. Imprinting status of gene loci

This method can detect fetal DNA molecules in maternal plasma that share the same genotype as the mother but have different epigenetic markers (LLM Poon et al., 2002 Clin Chem; 48: 35-41). To demonstrate the sensitivity of sequencing methods in selecting fetal DNA molecules in maternal plasma, the same strategy was applied to detect imprinted fetal alleles in maternal plasma samples. Two genomic imprinted regions were identified: H19 (chr11: 1,977,419-1,977,821, NCBI Build36/hg18) and MEST (chr7: 129,917,976-129,920,347, NCBI Build36/hg18). Both contained informative SNPs that distinguish maternal and fetal sequences. For H19 (a maternally expressed gene), for SNP rs2071094 (chr11:1,977,740) in the region, the mother is homozygous (A/A) and the fetus is heterozygous (A/C). One of the maternal A alleles is fully methylated while the others are unmethylated. However, in the placenta, the A allele is unmethylated, while the paternally inherited C allele is fully methylated. Two methylated reads with the C genotype were detected in maternal plasma, corresponding to imprinted paternal alleles from the placenta.

MEST, also known as PEG1, is a paternally expressed gene. For the SNP rs2301335 (chr7:129,920,062) at the imprinted locus, both mother and fetus are heterozygous (A/G). The G allele is methylated in maternal blood, while the A allele is unmethylated. The methylation pattern in the placenta is reversed, with the maternal A allele methylated and the paternal G allele unmethylated. The three unmethylated G alleles from the father are detectable in maternal plasma. In contrast, VAV1, a non-imprinted locus on chromosome 19 (chr19:6,723,621-6,724,121), does not exhibit any allele methylation pattern in tissue or plasma DNA samples.

Therefore, methylation status can be used to determine which DNA fragments originated from the fetus. For example, when the mother is GA heterozygous, the presence of only the A allele in maternal plasma cannot be used as a fetal marker. However, if the methylation status of A molecules in the plasma is distinguished, then methylated A molecules are fetal-specific, while unmethylated A molecules are maternally-specific, or vice versa.

The next focus was on loci that had been reported to have genomic imprinted in placental tissue. Based on the list of loci reported by Woodfine et al. (2011 Epigenetics Chromatin; 4:1), loci containing SNPs within the imprinted control region were further sorted. Four loci met the criteria and were H19, KCNQ10T1, MEST, and NESP.

Regarding the H19 and KCNQ10T1 readings in maternal blood cell samples, the maternal readings were SNP homozygous and showed approximately equal proportions of methylated and unmethylated readings. CVS and full-term placental tissue samples revealed that the fetus was heterozygous for both loci and each allele was exclusively methylated or unmethylated, exhibiting monoallelic methylation. In maternal plasma samples, paternally inherited fetal DNA molecules were detected for both loci. For H19, paternally inherited molecules were represented by sequencing reads containing fetal-specific alleles and being methylated. For KCNQ10T1, paternally inherited molecules were represented by sequencing reads containing fetal-specific alleles and being unmethylated.

On the other hand, for both MEST and NESP, the mother is heterozygous. For MEST, both the mother and fetus are GA heterozygous for the SNP. However, as is evident from the Watson chain data of maternal blood cells and placental tissue, the methylation status of CpG near the SNP is reversed in the mother and fetus. The A allele is unmethylated in the mother's DNA but methylated in the fetal DNA. For MEST, the maternal allele is methylated. Therefore, it can be indicated that the fetus has inherited the A allele from its mother (methylated in CVS) and the mother has inherited the A allele from its father (unmethylated in maternal blood cells). Interestingly, in maternal plasma samples, all four groups of molecules can be easily distinguished, including each of the mother's two alleles and each of the fetus's two alleles. Therefore, by combining the genotype information of imprinted loci with methylation status, it is easy to distinguish between maternally inherited fetal DNA molecules and background maternal DNA molecules (LLM Poon et al. 2002 Clin Chem; 48:35-41).

This method can be used to detect uniparental disomy. For example, if the father of the fetus is known to be homozygous for the G allele, the absence of unmethylated G alleles in the maternal plasma indicates a lack of paternal allele contribution. Furthermore, in such cases, the presence of both methylated G and methylated A alleles in the plasma of the pregnancy indicates maternal disomy, meaning the fetus has inherited two distinct alleles from the mother and none from the father. Alternatively, if both methylated A alleles (the fetal allele inherited from the mother) and unmethylated A alleles (the maternal allele inherited from the maternal grandfather) are detected in the maternal plasma, but the absence of unmethylated G alleles (the paternal allele that should have been inherited by the fetus) indicates maternal disomy, meaning the fetus has inherited two identical alleles from the mother and none from the father.

For NESP, the mother is heterozygous for GA in the SNP, while the fetus is homozygous for the G allele. For NESP, the paternal allele is methylated. In maternal plasma samples, the methylated paternally inherited fetal G allele can be easily distinguished from the unmethylated background maternal G allele.

VIII. Cancer/Donor

Some embodiments can be used to detect, screen, monitor (e.g., recurrence, remission, or response to treatment (e.g., presence or absence)), stage, classify (e.g., to help select the most appropriate treatment modality) and prognosis of cancer using circulating plasma/serum DNA methylation analysis.

Cancer DNA is known to exhibit abnormal DNA methylation (Herman et al. 2003 N Engl J Med; 349:2042-2054). For example, compared to non-cancer cells, CpG island promoters of genes such as tumor suppressor genes are highly methylated, while CpG sites in the genome are lowly methylated. Assuming that the methylation pattern of cancer cells can be reflected by the methylation pattern of tumor-derived plasma DNA molecules using the methods described herein, it is expected that the total methylation pattern in plasma will differ between individuals with cancer, whether compared to healthy individuals without cancer or to individuals whose cancer has been cured. The type of difference in methylation pattern can be based on quantitative differences in methylation density of the genome and/or methylation density of genomic segments. For example, because DNA from cancerous tissue is generally hypomethylated (Gama-Sosa et al., 1983 Nucleic Acids Res; 11:6883-6894), a decrease in plasma methylome or methylation density of genomic segments will be observed in the plasma of cancer patients.

Qualitative changes in methylation patterns should also be reflected in plasma methylome data. For example, plasma DNA molecules derived from genes that are hypermethylated only in cancer cells will show hypermethylation in the plasma of cancer patients when compared to plasma DNA molecules derived from the same gene but in samples from healthy controls. Because aberrant methylation occurs in most cancers, the methods described herein can be applied to detect all forms of malignancy with aberrant methylation, such as (but not limited to) malignancies of the lung, breast, colorectal, prostate, nasopharynx, stomach, testes, skin, nervous system, bone, ovary, liver, blood tissue, pancreas, uterus, kidney, bladder, lymphoid tissue, etc. Malignancies can have various histological subtypes, such as carcinoma, adenocarcinoma, sarcoma, fibroadenocarcinoma, neuroendocrine, and undifferentiated carcinoma.

On the other hand, it is expected that tumor-derived DNA molecules can be distinguished from background non-tumor-derived DNA molecules because the overall short size morphology of tumor-derived DNA is prominent for DNA molecules originating from loci with tumor-associated aberrant hypomethylation, which will have an additional impact on the size of the DNA molecule. Furthermore, several characteristic features associated with tumor DNA can be used to distinguish tumor-derived plasma DNA molecules from background non-tumor-derived plasma DNA molecules. These features include (but are not limited to) single nucleotide variants, copy number increases and losses, translocations, inversions, aberrant high or low methylation, and size morphology analysis. Because all these changes can occur independently, the combined use of these features can provide additional advantages for the sensitive and specific detection of cancer DNA in plasma.

A. Size and Cancer

The size of tumor-derived DNA molecules in plasma is similar to that of mononuclear body units and is shorter than the background non-tumor-derived DNA molecules coexisting in the plasma of cancer patients. Size parameters have been shown to be relevant to cancer, as described in U.S. Patent Application 13/789,553, which is incorporated herein by reference for all purposes.

Because both fetal and maternal DNA in blood plasma exhibit a relationship between molecular size and methylation status, tumor-derived DNA molecules are expected to show the same tendency. For example, in the plasma of cancer patients or in individuals screened for cancer, hypomethylated molecules will be shorter than hypermethylated molecules.

B. Methylation density in different tissues of cancer patients

In this example, plasma and tissue samples from a patient with hepatocellular carcinoma (HCC) were analyzed. Blood samples were collected from HCC patients before and one week after surgical resection of the tumor. Plasma and leukocyte layers were harvested after centrifugation of the blood samples. The resected tumor and adjacent non-tumor liver tissue were collected. DNA samples extracted from plasma and tissue samples were analyzed using massively parallel sequencing with and without prior bisulfite treatment. Plasma DNA from four healthy individuals without cancer was also analyzed as a control. Bisulfite treatment of the DNA samples converted unmethylated cytosine residues to uracil. These uracil residues will behave like thymidine in downstream polymerase chain reaction and sequencing. On the other hand, bisulfite treatment did not convert methylated cytosine residues to uracil. Following massively parallel sequencing, the sequencing reads were analyzed using Methy-Pipe (Jiang et al. Methy-Pipe: An integrated bioinformatics data analysis pipeline for whole genome methylome analysis, paper presented at the IEEE International Conference on Bioinformatics and Biomedicine Workshops, Hong Kong, 18 to 21 December 2010) to determine the methylation status of all CG dinucleotide positions, i.e., cytosine residues at CpG sites.

Figure 21A is Table 2100, showing the methylation density of plasma and tissue samples from HCC patients before surgery. CpG methylation density in relevant regions (e.g., CpG sites, promoters, or repeat regions) refers to the proportion of readings showing CpG methylation on CpG dinucleotides covering the genome out of the total readings. Methylation densities were similar in the leukocyte layer and non-tumor liver tissue. Based on data from all autosomes, the total methylation density in tumor tissue was 25% lower than that in the leukocyte layer and non-tumor liver tissue. Hypomethylation was consistent across all chromosomes. Plasma methylation density fell between values in non-malignant and cancerous tissues. This observation is consistent with the fact that both cancerous and non-cancer tissues contribute to circulating DNA in the peripheral blood of cancer patients. The hematopoietic system has been shown to be a major source of circulating DNA in individuals with inactive malignant symptoms (YYN Lui, et al. 2002 Clin Chem; 48:421-7). Therefore, plasma samples obtained from four healthy controls were also analyzed. Table 2150 in Figure 21B shows the number of sequence reads and sequencing depth achieved for each sample.

Figure 22 is a summary of Table 220, showing that the methylation density of autosomes in plasma samples from healthy controls ranged from 71.2% to 72.5%. These data illustrate the expected level of DNA methylation in plasma samples obtained from individuals without tumor DNA origin. In cancer patients, tumor tissue also releases DNA into circulation (KCA Chan et al. 2013 Clin Chem; 59:211-224; Liri et al. 2012 Science Translational Medicine; 4:162ra154). The presence of both tumor and non-tumor-derived DNA in the preoperative plasma of patients, attributed to the hypomethylation of HCC tumors, caused a decrease in methylation density compared to healthy control plasma. In fact, the methylation density of preoperative plasma samples fell between that of tumor tissue and healthy control plasma. This is because the methylation level of plasma DNA in cancer patients is influenced by the degree of aberrant methylation (hypomethylation in this case) in tumor tissue and the percentage concentration of tumor-derived DNA in circulation. Lower methylation density in tumor tissue and a higher percentage concentration of tumor-derived DNA in circulation result in lower plasma DNA methylation density in cancer patients. Most tumors have been reported to exhibit overall hypomethylation (Herman et al. 2003, New England Journal of Medicine; 349: 2042-2054; M. Sosa et al. 1983, Nucleic Acid Research; 11: 6883-6894). Therefore, current observations in HCC samples should also apply to other types of tumors.

In one embodiment, when the methylation level of tumor tissue is known, the methylation density of plasma DNA can be used to determine the percentage concentration of tumor-derived DNA in a plasma/serum sample. If a tumor sample is available or a tumor biopsy is available, then the methylation level of the tumor tissue, such as methylation density, can be obtained. In another embodiment, information about the methylation level of tumor tissue can be obtained from a study of methylation levels in a group of similar types of tumors, and this information (e.g., average or median level) is applied to a patient to be analyzed using the techniques described in this invention. The methylation level of tumor tissue can be determined by analyzing a patient's tumor tissue or inferred from analysis of tumor tissue from other patients with the same or similar types of cancer. The methylation of tumor tissue can be determined using a range of methylation-identifying platforms, including (but not limited to) massively parallel sequencing, single-molecule sequencing, microarrays (e.g., oligonucleotide arrays), or mass spectrometry (e.g., Sequenom, Inc.'s Epityper assay). In some embodiments, such analyses can be procedures sensitive to the methylation state of DNA molecules, including (but not limited to) cytosine immunoprecipitation and restriction enzyme digestion that detects methylation. When the methylation level of the tumor is known, the percentage concentration of tumor DNA in the plasma of cancer patients can be calculated after plasma methylome analysis.

The relationship between plasma methylation level P and fractional tumor DNA concentration f and tumor tissue methylation level TUM can be described as: P = BKG × (1-f) + TUM × f, where BKG is the background DNA methylation level in plasma derived from blood cells and other internal organs. For example, the total methylation density of all autosomes in a tumor biopsy tissue obtained from this HCC patient is 42.9%, which is the TUM value in this case. The average methylation density of plasma samples from four healthy controls is 71.6%, which is the BKG value in this case. The plasma methylation density before surgery is 59.7%. Using these values, f is estimated to be 41.5%.

In another embodiment, when the percentage concentration of tumor-derived DNA in a plasma sample is known, the methylation level of tumor tissue can be noninvasively assessed based on plasma methylome data. The percentage concentration of tumor-derived DNA in a plasma sample can be determined by other genetic analyses, such as genome-wide analysis of allele loss (GAAL) and single nucleotide mutation analysis as previously described (US Patent Application 13/308,473; Chan et al. 2013 Clin Chem; 59:211-24). This calculation is based on the same relationship as described above, except that in this embodiment, the value of f is known, while the value of TUM becomes unknown. Inferences can be made against the whole genome or a portion of the genome, similar to the data observed in determining placental tissue methylation levels from maternal plasma data.

In another embodiment, variations or patterns in methylation density between regions can be used to distinguish individuals with cancer from those without. The resolution of methylation analysis can be further increased by dividing the genome into regions of a specific size (e.g., 1 Mb). In such embodiments, the methylation density of each 1 Mb region is calculated for collected samples, such as leukocyte layers, excised HCC tissue, non-tumor liver tissue near the tumor, and plasma collected before and after tumor resection. In another embodiment, the region size does not need to be constant. In one implementation, the number of CpG sites within each region remains constant, while the size of the region itself can vary.

Figures 23A and 23B show the methylation density of leukocyte layer, tumor tissue, non-tumor liver tissue, preoperative plasma, and postoperative plasma in HCC patients. Figure 23A is Figure 2300, corresponding to the results for chromosome 1. Figure 23B is Figure 2350, corresponding to the results for chromosome 2.

For most of the 1Mb window, the methylation density is similar in the leukocyte layer and in non-tumor liver tissue near the tumor, while the methylation density in tumor tissue is lower. Preoperative plasma methylation density falls between that of tumor and non-malignant tissue. The methylation density of the queried genomic regions in tumor tissue can be inferred using preoperative plasma methylation data and fractional tumor DNA concentration. The method is the same as described above, using methylation density values for all autosomes. The tumor methylation can also be inferred using this higher resolution methylation data from plasma DNA. Other region sizes, such as 300kb, 500kb, 2Mb, 3Mb, 5Mb, or greater than 5Mb, can also be used. In one embodiment, the region size does not need to be constant. In one implementation, the number of CpG sites within each region remains constant, while the size of the region itself can vary.

C. Comparison of plasma methylation density between cancer patients and healthy individuals

As shown in 2100, the methylation density of plasma DNA in cancer patients before surgery is lower than that in non-malignant tissues. This is likely due to the presence of hypomethylated DNA from tumor tissue. This lower plasma DNA methylation density could potentially be used as a biomarker for detecting and monitoring cancer. For cancer surveillance, if cancer progresses, the amount of cancer-derived DNA in the plasma increases over time. In this example, the increased amount of circulating cancer-derived DNA in the plasma will cause a further decrease in plasma DNA methylation density at the whole-genome level.

Conversely, if cancer responds to treatment, the amount of cancer-derived DNA in the plasma will decrease over time. In this example, the decrease in the amount of cancer-derived DNA in the plasma will cause an increase in plasma DNA methylation density. For instance, if a lung cancer patient with an epidermal growth factor receptor mutation has been treated with targeted therapy such as tyrosine kinase inhibition, an increase in plasma DNA methylation density will indicate a response to treatment. Subsequently, the emergence of tumor clones resistant to tyrosine kinase inhibition will be associated with a decrease in plasma DNA methylation density, indicating recurrence.

Plasma methylation density measurements can be performed continuously, and the rate of change in such measurements can be calculated and used to predict or correlate clinical progression, remission, or prognosis. For selected loci that are hypermethylated in cancerous tissues but hypomethylated in normal tissues, such as promoter regions of numerous tumor suppressor genes, the relationship between cancer progression and favorable response to treatment will be the opposite of the above pattern.

To demonstrate the feasibility of this method, the DNA methylation density of plasma samples collected from cancer patients before and after surgical removal of the tumor will be compared with that of plasma DNA obtained from four healthy control individuals.

Table 2200 shows the DNA methylation density and combined values for each autosome in preoperative and postoperative plasma samples from cancer patients and four healthy controls. For all chromosomes, the methylation density of preoperative plasma DNA samples was lower than that of postoperative samples and plasma samples from the four healthy individuals. The difference in plasma DNA methylation density between preoperative and postoperative samples provides supporting evidence that the lower methylation density in preoperative plasma samples is due to the presence of DNA from HCC tumors.

The reversal of DNA methylation density in postoperative plasma samples to levels similar to those in healthy controls indicates that much of the tumor-derived DNA has been eliminated due to surgical removal of the source, the tumor. These data suggest that using data available from large genomic regions (e.g., all autosomes or individual chromosomes) to determine that preoperative plasma methylation density has lower levels than healthy controls allows for the identification, i.e., diagnostic or screening tests for cases of cancer.

Preoperative plasma data also showed significantly lower methylation levels than postoperative plasma, indicating that plasma methylation levels can also be used to monitor tumor burden, thus predicting and monitoring cancer progression in patients. Reference values can be determined from the plasma of healthy controls or individuals at risk of cancer but currently cancer-free. Individuals at risk of HCC include those with chronic hepatitis B or C infection, those with hemochromatosis, and those with cirrhosis.

The presence of tumor DNA in the plasma of non-pregnant individuals can be assessed using plasma methylation density values exceeding, for example, below, a threshold defined based on reference values. To detect the presence of hypomethylated circulating tumor DNA, the threshold can be defined as being below the 5th or 1st percentile of values in the control population, or based on the number of standard deviations, such as being 2 or 3 standard deviations (SD) below the mean methylation density value of the control population, or based on a determined median fold (MoM). For hypermethylated tumor DNA, the threshold can be defined as being above the 95th or 99th percentile of values in the control population, or based on the number of standard deviations, such as exceeding the mean methylation density value of the control population by 2 or 3 SD, or based on a determined median fold (MoM). In one embodiment, the control population is age-matched to the test individuals. Age matching does not need to be accurate and can be performed within age groups (e.g., 30 to 40 years old, for a 35-year-old test individual).

Next, the methylation density of the 1Mb region was compared between plasma samples from cancer patients and four control individuals. For illustration, the results for chromosome 1 are shown.

Figure 24A is Figure 2400, showing the methylation density of plasma from HCC patients before surgery. Figure 24B is Figure 2450, showing the methylation density of plasma from HCC patients after surgery. Blue dots represent results from control individuals, and red dots represent results from plasma samples from HCC patients.

As shown in Figure 24A, for most regions, the methylation density of preoperative plasma from HCC patients was lower than that of controls. Similar patterns were observed on other chromosomes. As shown in Figure 24B, for most regions, the methylation density of postoperative plasma from HCC patients was similar to that of controls. Similar patterns were observed on other chromosomes.

To assess whether a test individual has cancer, the individual's results are compared to those of a reference group. In one embodiment, the reference group may consist of a large number of healthy individuals. In another embodiment, the reference group may consist of individuals with non-malignant conditions such as chronic hepatitis B infection or cirrhosis. The difference in methylation density between the individual and the reference group can then be quantitatively tested.

In one embodiment, the reference range may be derived from values in a control group. The deviation of the test individual's result from the upper or lower limit of the reference group can then be used to determine whether the individual has a tumor. This quantity will be affected by the percentage concentration of tumor-derived DNA in the plasma and the difference in methylation levels between malignant and non-malignant tissues. A higher percentage concentration of tumor-derived DNA in the plasma will cause a greater difference in methylation density between the test plasma sample and the control. A greater difference in methylation levels between malignant and non-malignant tissues will also cause a greater difference in methylation density between the test plasma sample and the control. In yet another embodiment, different reference groups are selected for test individuals of different age ranges.

In another embodiment, for each 1Mb region, the mean and SD of methylation density for four control individuals are calculated. Then, for the corresponding region, the difference between the methylation density of the HCC patient and the mean of the control individuals is calculated. In one embodiment, this difference is then divided by the SD of the corresponding region to determine the z-score. In other words, the z-score represents the difference in methylation density between the tested and control plasma samples, expressed as the number of SDs relative to the mean of the control individuals. A z-score > 3 for a region indicates that the plasma DNA of the HCC patient in this region is more than 3 SD higher than that of the control individuals, while a z-score < -3 for a region indicates that the plasma DNA of the HCC patient in this region is more than 3 SD lower than that of the control individuals.

Figures 25A and 25B show the z-scores of plasma DNA methylation density in plasma samples from HCC patients before surgery (curve 2500) and after surgery (curve 2550), using plasma methylome data from four healthy control individuals as a reference for chromosome 1. Each point represents the result for a 1 Mb region. Black dots indicate regions with z-scores between -3 and 3. Red dots indicate regions with z-scores < -3.

Figure 26A is from Table 2600, showing the z-score data for plasma before and after surgery. In preoperative plasma samples, the z-score for most regions of chromosome 1 (80.9%) was <-3, indicating significantly lower methylation of plasma DNA in HCC patients compared to controls. Conversely, the number of red dots in postoperative plasma samples was substantially reduced (8.3% across all regions of chromosome 1), indicating that most of the tumor DNA, attributable to the surgical removal of the source of tumor DNA in the peripheral blood circulation, had been removed from the peripheral blood circulation.

Figure 26B is Circos plot 2620, showing the z-scores of plasma DNA methylation density in plasma samples from HCC patients before and after surgery, using four healthy controls as references, for all 1Mb regions analyzed from all autosomes. The outermost ring shows the G-banding map of human autosomes. The middle ring shows the data from the preoperative plasma samples. The innermost ring shows the data from the postoperative plasma samples. Each dot represents the result for a 1Mb region. Black dots represent regions with z-scores between -3 and 3. Red dots represent regions with z-scores < -3. Green dots represent regions with z-scores > 3.

Figure 26C is Table 2640, showing the distribution of z-scores for all 1Mb regions of the whole genome in preoperative and postoperative plasma samples from HCC patients. The results indicate that for most regions of the whole genome (85.2% of all 1Mb regions), preoperative plasma DNA from HCC patients was more hypomethylated than in controls. Conversely, most regions in postoperative plasma samples (93.5% of all 1Mb regions) did not show significant hypermethylation or hypomethylation compared to controls. These data suggest that for this HCC, substantially predominantly hypomethylated tumor DNA was no longer present in the postoperative plasma samples.

In one embodiment, the number, percentage, or proportion of regions with a z-score <-3 can be used to indicate the presence of cancer. For example, as shown in Table 2640, 2330 (85.2%) of the 2734 regions analyzed in preoperative plasma showed a z-score <-3, while only 171 (6.3%) of the 2734 regions analyzed in postoperative plasma showed a z-score <-3. The data indicate a significantly higher tumor DNA load in preoperative plasma than in postoperative plasma.

The threshold for the number of regions can be determined using statistical methods. For example, based on a normal distribution, it is expected that approximately 0.15% of the regions will have a z-score <-3. Therefore, the cutoff number of regions could be 0.15% of the total number of regions analyzed. In other words, if more than 0.15% of the regions in a plasma sample from a non-pregnant individual show a z-score <-3, then there is a source of hypomethylated DNA in the plasma, i.e., cancer. For example, in this instance, 0.15% of the 2734 1Mb regions used for analysis is approximately 4 regions. Using this value as a threshold, both pre- and post-operative plasma samples contained hypomethylated tumor-derived DNA, although the amount was significantly higher in the pre-operative plasma sample than in the post-operative plasma sample. For the four healthy control individuals, no regions showed significant hypermethylation or hypomethylation. Other thresholds (e.g., 1.1%) can be used and can be varied depending on the needs of the analysis used. As another example, the cutoff percentage can be varied based on the statistical distribution and the required sensitivity and acceptable specificity.

In another embodiment, the cutoff number can be determined by analyzing a large number of cancer patients and individuals without cancer using receiver operating characteristic (ROC) curve analysis. To further validate the specificity of this method, plasma samples from patients seeking medical advice regarding non-malignant conditions (C06) were analyzed. A region of 1.1% showed a z-score <-3. In one embodiment, different thresholds can be used to classify different levels of disease states. Lower percentage thresholds can be used to distinguish between healthy states and benign conditions, while higher percentage thresholds can be used to distinguish between benign conditions and malignant conditions.

The diagnostic performance of plasma hypomethylation analysis using massively parallel sequencing appears to be superior to polymerase chain reaction (PCR)-based amplification using specific classes of repeating elements (e.g., long dispersed nuclear element-1 (LINE-1)) (P Tangkijvanich et al. 2007 Clin Chim Acta; 379:127-133). One possible explanation for this observation is that while hypomethylation is prevalent in the tumor genome, it exhibits a degree of heterogeneity from one genomic region to the next.

In fact, the average plasma methylation density of the reference individuals was observed across genomic variations (Figure 56). Each red dot in Figure 56 represents the average methylation density of a 1Mb region in 32 healthy individuals. The graph shows all 1Mb regions across the genome analysis. The number within each box represents the chromosome number. The average methylation density was observed to vary across regions.

Simple PCR-based analyses cannot account for this heterogeneity between regions in their diagnostic algorithms. This heterogeneity will broaden the range of methylation density observed in healthy individuals. Subsequently, for samples considered to exhibit hypomethylation, a greater reduction in methylation density will be required. This will lead to decreased test sensitivity.

In contrast, methods based on massively parallel sequencing divide the genome into 1Mb regions (or other sizes) and measure the methylation density of such regions individually. This approach reduces the impact of variations in baseline methylation density across different genomic regions when comparing each region between test samples and controls. In fact, within the same region, the variation between individuals across 32 healthy controls was relatively small. Across 32 healthy controls, the coefficient of variation (CV) for 95% of the regions was ≤1.8%. However, to further enhance the sensitivity of detecting cancer-related hypomethylation, comparisons can be made across multiple genomic regions. Sensitivity will be enhanced by testing multiple genomic regions when a specific region in a cancer sample happens not to show hypomethylation when only one region is tested, as this will protect against the influence of biological variations.

Comparing the methylation density of equivalent genomic regions between control and test samples, and methods that perform this comparison on multiple genomic regions (e.g., testing each genomic region separately and then potentially combing through these results), offer a high signal-to-noise ratio for detecting cancer-associated hypomethylation. This massively parallel sequencing method is illustrated illustratively. Other methods that can determine the methylation density of multiple genomic regions and allow comparisons of the methylation density of corresponding regions between control and test samples can achieve similar results. For example, hybridization probes or molecular inversion probes that can target plasma DNA molecules derived from specific genomic regions and determine the methylation level of that region can be designed to achieve the desired effect.

In another embodiment, the sum of z-scores across all regions can be used to determine the presence of cancer or to monitor continuous changes in plasma DNA methylation levels. Due to the overall hypomethylation of tumor DNA, the sum of z-scores in plasma collected from individuals with cancer is lower than in healthy controls. The sum of z-scores in plasma from HCC patients before and after surgery were -49843.8 and -3132.13, respectively.

In other embodiments, other methods can be used to study the methylation level of plasma DNA. For example, mass spectrometry (ML Chen et al. 2013 Clin Chem; 59:824-832) or massively parallel sequencing can be used to determine the proportion of methylated cytosine residues to the total cytosine residues. However, because most cytosine residues are not in a CpG dinucleotide background, the proportion of methylated cytosine in the total cytosine residues will be relatively small compared to the methylation levels assessed in the CpG dinucleotide context. The methylation levels of tissue and plasma samples obtained from HCC patients and four plasma samples obtained from healthy controls were determined. Methylation levels were measured using whole-genome massively parallel sequencing data in CpG, any cytosine background, and CHG and CHH backgrounds. H refers to adenine, thymine, or cytosine residues.

Figure 26D is Table 2660, showing the methylation levels of tumor tissue and preoperative plasma samples overlapping with some control plasma samples when using CHH and CHG backgrounds. When compared with leukocyte layer, non-tumor liver tissue, postoperative plasma samples, and healthy control plasma samples, the methylation levels in both tumor tissue and preoperative plasma samples were consistently lower in both CpG and unspecified cytosine. However, data based on methylated CpG, i.e., methylation density, show a wider dynamic range than data based on methylated cytosine.

In other embodiments, the methylation status of plasma DNA can be determined using methods employing antibodies against methylated cytosine, such as methylated DNA immunoprecipitation (MeDIP). However, these methods are not expected to be as accurate as sequencing-based methods because antibody binding is variable. In yet another embodiment, the level of 5-hydroxymethylcytosine in plasma DNA can be determined. In this regard, reduced levels of 5-hydroxymethylcytosine have been found to be an epigenetic characteristic of certain cancers, such as melanoma (C.Lian et al. 2012 Cell; 150:1135-1146).

In addition to HCC, the applicability of this method to other types of cancer was investigated. Plasma samples from two patients with lung adenocarcinoma (CL1 and CL2), two patients with nasopharyngeal carcinoma (NPC1 and NPC2), two patients with colorectal cancer (CRC1 and CRC2), one patient with metastatic neuroendocrine tumor (NE1), and one patient with metastatic leiomyosarcoma (SMS1) were analyzed. The plasma DNA of these individuals was bisulfite-converted and sequenced at one end (50 bp) using the ilumina HiSeq2000 platform. Four healthy controls were used as a reference population for analyzing these eight patients. The 50 bp sequence reads were used. The entire genome was divided into 1 Mb regions. Using data from the reference population, the mean methylation density and SD of each region were calculated. The results for the eight cancer patients were then expressed as z-scores, representing the number of SDs from the reference population mean. A positive value indicates that the methylation density of the test case was lower than the reference population mean, and vice versa. Table 2780 in Figure 27I shows the number of sequence reads and sequencing depth achieved for each sample.

Figures 27A-H show Circos plots of methylation density for eight cancer patients according to an embodiment of the present invention. Each dot represents a 1 Mb region. Black dots represent regions with z-scores between -3 and 3. Red dots represent regions with z-scores < -3. Green dots represent regions with z-scores > 3. The interval between two consecutive lines represents a z-score difference of 20.

For most types of cancer, including lung cancer, nasopharyngeal carcinoma, colorectal cancer, and metastatic neuroendocrine tumors, significant hypomethylation was observed in multiple regions across the genome. Interestingly, in addition to hypomethylation, significant hypermethylation was also observed in multiple regions across the genome in the case of metastatic leiomyosarcoma. Leiomyosarcoma has an embryonic origin from the mesoderm, while the other cancer types in the remaining seven patients had an embryonic origin from the ectoderm. Therefore, the DNA methylation pattern of sarcoma may differ from that of carcinoma.

This case demonstrates that plasma DNA methylation patterns can also be used to differentiate between different types of cancer, in this example, carcinoma and sarcoma. These data also indicate that the described method can be used to detect abnormally high methylation associated with malignant diseases. For all eight cases, only plasma samples were obtained, and tumor tissue was not analyzed. This demonstrates that even without prior knowledge of tumor tissue methylation patterns or levels, tumor-derived DNA in plasma can be easily detected using the described method.

Figure 27J is a direct translation of Table 2790, showing the distribution of z-scores for all 1Mb regions of the whole genome in the plasma of patients with different malignancies. It shows the percentage of regions with z-scores <-3, -3 to 3, and >3 for each case. More than 5% of the regions in all cases showed z-scores <-3. Therefore, if 5% of the regions are used as a threshold for significant hypomethylation to classify samples as cancer-positive, then all these cases would be classified as cancer-positive. The results demonstrate that hypomethylation is likely a common phenomenon across different types of cancer, and that plasma methylome analysis would be applicable for detecting different types of cancer.

D. Method

Figure 28 is a flowchart of method 2800, which, according to an embodiment of the invention, analyzes a biological sample of an organism to determine the classification of cancer grade. The biological sample includes DNA derived from normal cells and may include DNA from cancer-associated cells. At least some of the DNA in the biological sample may be free.

At box 2810, multiple DNA molecules from a biological sample are analyzed. The analysis of DNA molecules may include determining the location of DNA molecules within the organism's genome and determining whether the DNA molecules are methylated at one or more sites. The analysis can be performed by receiving sequence reads from sequencing that identifies methylation, thus the analysis can be performed solely on data previously obtained from the DNA. In other embodiments, the analysis may include actual sequencing or other steps to obtain data.

At box 2820, for each of the plurality of sites, the corresponding number of DNA molecules methylated at that site is determined. In one embodiment, the site is a CpG site, and may be only certain CpG sites, selected using one or more criteria mentioned herein. Once normalized using the total number of DNA molecules analyzed at a particular site, such as the total number of sequence reads, the number of methylated DNA is equivalent to determining the number of unmethylated DNA. For example, an increase in the CpG methylation density of a region is equivalent to a decrease in the density of unmethylated CpG in the same region.

At box 2830, a first methylation level is calculated based on the corresponding number of DNA molecules methylated at multiple sites. The first methylation level may correspond to a methylation density determined based on the number of DNA molecules corresponding to the multiple sites. A site may correspond to multiple loci or only one locus.

At box 2840, a first methylation level is compared to a first threshold. The first threshold can be a reference methylation level or related to a reference methylation level (e.g., a specified distance from a normal level). The reference methylation level can be determined from a sample from a cancer-free individual or from a locus or organism known not to be associated with cancer in the organism. The first threshold can be established from a reference methylation level determined from a previously obtained biological sample of an organism, which may precede the biological sample being tested.

In one embodiment, the first threshold is a specified distance (e.g., a specified number of standard deviations) from a reference methylation level established from a biological sample obtained from a healthy organism. Comparisons can be made by determining the difference between the first methylation level and the reference methylation level, and subsequently comparing the difference to a threshold corresponding to the first threshold (e.g., to determine if there is a statistically significant difference between the methylation level and the reference methylation level).

At box 2850, a cancer grade classification is determined based on comparison. Examples of cancer grades include whether an individual has cancer or precancerous symptoms, or an increased likelihood of developing cancer. In one embodiment, a first threshold may be determined from a sample previously obtained from the individual (e.g., a reference methylation level may be determined from a previous sample).

In some embodiments, a first methylation level may correspond to the number of regions whose methylation levels exceed a threshold. For example, multiple regions of an organism's genome may be identified. These regions may be identified using criteria mentioned herein, such as a certain length or a certain number of sites. One or more sites (e.g., CpG sites) may be identified within each region. A regional methylation level may be calculated for each region. The first methylation level is for a first region. The methylation level of each region is compared to a corresponding regional threshold, which may be the same or vary with the region. The regional threshold for the first region is the first threshold. The corresponding regional threshold may be a specified amount of distance from a reference methylation level (e.g., 0.5), thereby counting only regions that differ significantly from the reference, which may be determined from a non-cancer individual.

A first number of regions whose methylation levels exceed a corresponding region threshold can be determined and compared with the threshold to determine the classification. In one implementation, the threshold is a percentage. The comparison of the first number with the threshold may include dividing the first number of regions by a second number of regions (e.g., all regions) before comparing with the threshold, for example as part of a normalization process.

As mentioned above, the percentage concentration of tumor DNA in a biological sample can be used to calculate a first threshold. A percentage concentration can be simply assessed as exceeding a minimum value, while samples with percentage concentrations below the minimum value can be flagged as, for example, unsuitable for analysis. The minimum value can be determined based on the expected difference between the tumor methylation level and a reference methylation level. For example, if the difference is 0.5 (e.g., as a threshold), then a sufficiently high tumor concentration would be required to satisfy this difference.

Specific techniques from Method 1300 can be used in Method 2800. In Method 1300, copy number variations can be identified for tumors (e.g., a first chromosomal region of the tumor can be tested for copy number changes relative to a second chromosomal region of the tumor). Therefore, Method 1300 can presuppose the presence of a tumor. In Method 2800, a sample can be tested for any indication of the presence of a tumor, regardless of any copy number characteristics. Some techniques of the two methods may be similar. However, Method 2800's thresholds and methylation parameters (e.g., normalized methylation levels) can detect statistical differences from a reference methylation level to a mixture of cancer DNA and non-cancer DNA, where some regions may have copy number variations, relative to a reference methylation level. Therefore, the reference value of Method 2800 can be determined from cancer-free samples, such as from a cancer-free organism or from non-cancer tissue from the same patient (e.g., previously collected plasma or from a simultaneously obtained sample known to be cancer-free (which can be determined from cellular DNA)).

E. Predicting the minimum percentage concentration of tumor DNA to be detected using plasma DNA methylation analysis.

One approach to measuring the sensitivity of a method for detecting cancer using plasma DNA methylation levels involves the minimum fraction of tumor-derived DNA concentration required to reveal changes in plasma DNA methylation levels compared to a control. Test sensitivity also depends on the degree of difference between DNA methylation in tumor tissue and baseline plasma DNA methylation levels in healthy controls or blood cells. Blood cells are the primary source of DNA in the plasma of healthy individuals. A greater difference makes it easier to distinguish cancer patients from non-cancer individuals and results in a lower detection limit reflecting tumor origin in plasma, as well as higher clinical sensitivity in detecting cancer patients. Furthermore, variations in plasma DNA methylation in healthy individuals or individuals of different ages (G Hannum et al. 2013 Mol Cell; 49:359-367) also affect the sensitivity for detecting methylation changes associated with the presence of cancer. Smaller changes in plasma DNA methylation in healthy individuals will make it easier to detect changes caused by the presence of small amounts of cancer-derived DNA.

Figure 29A, curve 2900, illustrates the distribution of methylation density in the reference individuals, assuming this distribution follows a normal distribution. This analysis is based on providing only one methylation density value per plasma sample, such as the methylation density of all autosomes or a specific chromosome. It illustrates how the specificity of the analysis will be affected. In one embodiment, a threshold of 3 SD lower than the average DNA methylation density of the reference individuals is used to determine whether the test sample is significantly lower in methylation than the sample from the reference individuals. When using this threshold, approximately 0.15% of non-cancer individuals are expected to have false positive results classified as having cancer, producing a specificity of 99.85%.

Figure 29B, curve 2950, illustrates the distribution of methylation density in reference individuals and cancer patients. The threshold is 3 SD lower than the mean methylation density of the reference individuals. If the mean methylation density of cancer patients is 2 SD lower than the threshold (i.e., 5 SD lower than the mean of the reference individuals), then it would be expected that 97.5% of cancer individuals have a methylation density below the threshold. In other words, if a single methylation density value is provided for each individual, such as when analyzing the total methylation density of the entire genome across all autosomes or a specific chromosome, then the expected sensitivity would be 97.5%. The difference in mean methylation density between the two populations is influenced by two factors: the degree of difference in methylation levels between cancer and non-cancer tissues and the percentage concentration of tumor-derived DNA in the plasma sample. The higher these two parameters are, the greater the difference in methylation density values between the two populations. Furthermore, the lower the SD of the distribution of methylation density between the two populations, the less overlap there is in the distribution of methylation density between the two populations.

Here, we use a hypothetical example to illustrate this concept. Assume the methylation density of tumor tissue is approximately 0.45 and the methylation density of plasma DNA in a healthy individual is approximately 0.7. These hypothetical values are similar to those obtained from HCC patients, where the total methylation density of autosomes is 42.9% and the average methylation density of autosomes from plasma samples from healthy controls is 71.6%. Assuming the CV for measuring genome-wide plasma DNA methylation density is 1%, the threshold would be 0.7 × (100% - 3 × 1%) = 0.679. To achieve a sensitivity of 97.5%, the average methylation density of plasma DNA from cancer patients would need to be approximately 0.679 - 0.7 × (2 × 1%) = 0.665. Let f represent the percentage concentration of tumor-derived DNA in the plasma sample. Then f can be calculated as (0.7 - 0.45) × f = 0.7 - 0.665. Therefore, f is approximately 14%. Based on this calculation, it is estimated that if the total methylation density of the whole genome is used as a diagnostic parameter, the minimum percentage concentration that can be detected in plasma is 14% in order to achieve a diagnostic sensitivity of 97.5%.

The following analysis was performed on data obtained from HCC patients. For this purpose, methylation density was measured only once per sample, based on values assessed by all autosomes. The mean methylation density in plasma samples obtained from healthy individuals was 71.6%. The SD of methylation density for these four samples was 0.631%. Therefore, the threshold for plasma methylation density would need to be 71.6% - 3 × 0.631% = 69.7% to achieve a z-score < -3 and a specificity of 99.85%. To achieve a sensitivity of 97.5%, the mean plasma methylation density in cancer patients would need to be 2 SDs lower than the threshold, i.e., 68.4%. Because the methylation density in tumor tissue is 42.9% and using the formula: P = BKG × (1 - f) + TUM × f, f would need to be at least 11.1%.

In another embodiment, the methylation density of different genomic regions can be analyzed separately, for example, as shown in Figures 25A or 26B. In other words, multiple measurements of methylation levels are performed for each sample. As explained below, significant hypomethylation can be detected at much lower fractions of tumor DNA concentration in plasma, thus enhancing the diagnostic performance of plasma DNA methylation analysis for cancer detection. The number of genomic regions exhibiting significant deviations in methylation density from a reference population can be counted. The number of genomic regions can then be compared with a threshold to determine whether significant hypomethylation of plasma DNA exists across the population spanning the studied genomic regions, such as a 1Mb region across the entire genome. The threshold can be established by analyzing a group of cancer-free reference individuals or derived mathematically, for example, based on a normal distribution function.

Figure 30 shows curve 3000, illustrating the distribution of methylation density in plasma DNA of healthy individuals and cancer patients. The methylation density of each 1Mb region is compared to the corresponding value in the reference population. The percentage of regions exhibiting significant hypomethylation (3 SD lower than the reference population mean) is determined. A threshold of 10% significant hypomethylation is used to determine the presence of tumor-derived DNA in the plasma sample. Other thresholds, such as 5%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 60%, 70%, 80%, or 90%, can also be used depending on the desired test sensitivity and specificity.

For example, to classify a sample as containing tumor-derived DNA, a threshold of 10% of 1Mb regions showing significant hypomethylation (z-score < -3) can be used. If more than 10% of regions are significantly hypomethylated compared to the reference population, the sample is classified as positive for a cancer test. For each 1Mb region, a threshold of 3 SD lower than the average methylation density of the reference population is used to define a sample as significantly hypomethylated. For each 1Mb region, if the average plasma DNA methylation density of cancer patients is 1.72 SD lower than the average plasma DNA methylation density of reference individuals, there is a 10% chance that the methylation density value of any particular region in the cancer patient will be below the threshold (i.e., z-score < -3) and result in a positive result. Subsequently, if all 1Mb regions of the whole genome are examined, approximately 10% of regions will be expected to show a positive result with significantly lower methylation density (i.e., z-score < -3). Assuming the total methylation density of plasma DNA in a healthy individual is approximately 0.7 and the coefficient of variation (CV) for measuring plasma DNA methylation density per 1 Mb region is 1%, the average methylation density of plasma DNA in a cancer patient would need to be 0.7 × (100% - 1.72 × 1%) = 0.68796. f is the percentage concentration of tumor-derived DNA in the plasma to achieve this average plasma DNA methylation density. Assuming the methylation density of tumor tissue is 0.45, f can be calculated using the following equation:

Wherein represents the average methylation density of plasma DNA in the reference individual, M _{_tumor} represents the methylation density of tumor tissue in the cancer patient; and represents the average methylation density of plasma DNA in the cancer patient.

Using this equation, (0.7-0.45)×f＝0.7-0.68796. Therefore, the minimum percentage concentration that can be detected using this method will be inferred to be 4.8%. Sensitivity can be further enhanced by reducing the cutoff percentage of significantly lower methylated regions, for example from 10% to 5%.

As illustrated in the examples above, the sensitivity of this method is determined by the degree of difference in methylation levels between cancer and non-cancer tissues, such as blood cells. In one embodiment, only chromosomal regions exhibiting large differences in methylation density between plasma DNA from non-cancer individuals and tumor tissues are selected. In another embodiment, only regions with a methylation density difference > 0.5 are selected. In other embodiments, differences of 0.4, 0.6, 0.7, 0.8, or 0.9 can be used to select suitable regions. In yet another embodiment, the actual size of the genomic region is not fixed. In practice, genomic regions are defined, for example, based on a fixed read depth or a fixed number of CpG sites. For each sample, the methylation levels of multiple of these genomic regions are evaluated.

Figure 31, or Chart 3100, shows the distribution of the difference in methylation density between plasma DNA from healthy individuals and the mean methylation density in tumor tissue from HCC patients. Positive values indicate higher methylation density in plasma DNA from healthy individuals, while negative values indicate higher methylation density in tumor tissue.

In one embodiment, regions with the greatest difference in methylation density between cancerous and non-cancer tissues can be selected, such as regions with a difference > 0.5, regardless of whether these regions are hypomethylated or hypermethylated tumors. Assuming the percentage concentration of tumor-derived DNA in plasma is the same, the detection limit for the percentage concentration of tumor-derived DNA in plasma can be reduced by focusing on these regions because the distribution of plasma DNA methylation levels differs significantly between cancerous and non-cancer individuals. For example, if only regions with a difference > 0.5 are used and a threshold of 10% of regions being significantly hypomethylated is applied to determine whether a tested individual has cancer, then the minimum percentage concentration (f) of tumor-derived DNA detected can be calculated using the following equation: where represents the average methylation density of plasma DNA in the reference individual, M _{_tumor} represents the methylation density of tumor tissue in the cancer patient; and represents the average methylation density of plasma DNA in the cancer patient.

If the difference in methylation density between the plasma and tumor tissue of a reference individual is at least 0.5, then 0.5 × f = 0.7 - 0.68796 and f = 2.4%. Therefore, by focusing on regions with higher methylation density differences between cancerous and non-cancer tissues, the lower limit for fractional tumor-derived DNA can be reduced from 4.8% to 2.4%. Information on which regions will show a greater degree of methylation difference between cancer and non-cancer tissues such as blood cells can be determined from tumor tissues of the same organ or tissue type obtained from other individuals.

In another embodiment, the parameters can be derived from the methylation density of plasma DNA across all regions and take into account the difference in methylation density between cancerous and non-cancer tissues. Regions with larger differences can be assumed to have higher weights. In one embodiment, when calculating the final parameters, the difference in methylation density between cancerous and non-cancer tissues in each region can be directly used as the weight for that specific region.

In another embodiment, different types of cancer can have different methylation patterns in tumor tissue. Cancer-specific weighted morphologies can be derived from the degree of methylation in a particular type of cancer.

In another embodiment, the interregional relationship of methylation density can be determined in individuals with and without cancer. In Figure 8, it can be observed that in a small number of regions, tumor tissue is more methylated than the plasma DNA of the reference individual. Therefore, the extreme values of the differences can be selected, such as regions with differences > 0.5 and differences < 0. The ratio of methylation density in these regions can then be used to indicate whether the tested individual has cancer. In other embodiments, the difference and quotient of methylation density in different regions can be used as parameters indicating interregional relationships.

Further evaluation of the method was conducted using methylation density detection of multiple genomic regions, as illustrated by data obtained from HCC patients, to assess the detection sensitivity of tumors. First, readings from preoperative plasma were mixed with readings from plasma samples obtained from healthy controls to simulate plasma samples containing fractional tumor DNA concentrations ranging from 20% to 0.5%. Subsequently, the percentage of methylation density corresponding to 1Mb regions (out of 2,734 regions in the whole genome) with z-scores <-3 was scored. When the fractional tumor DNA concentration in plasma was 20%, 80.0% of the regions showed significant hypomethylation. The corresponding data for fractional tumor DNA concentrations in plasma of 10%, 5%, 2%, 1%, and 0.5% were 67.6%, 49.7%, 18.9%, 3.8%, and 0.77%, respectively, showing hypomethylation. Because the theoretical limit for the number of regions showing z-scores <-3 in the control sample is 0.15%, the data shows that even at a tumor percentage concentration of only 0.5%, there are still more regions (0.77%) that exceed the theoretical cutoff.

Figure 32A is a copy of Table 3200, showing the effect of reducing sequencing depth when plasma samples contain 5% or 2% tumor DNA. Even with an average sequencing depth of only 0.022 times that of the haploid genome, a high proportion of regions exhibiting significant hypomethylation (>0.15%) can still be detected.

Figure 32B, which is Figure 3250, shows the methylation density of repeating elements and non-repeating regions in plasma from four healthy controls, leukocyte layers from HCC patients, normal liver tissue, tumor tissue, pre-operative plasma, and post-operative plasma samples. It can be observed that repeating elements are more methylated (higher methylation density) than non-repeating regions in both cancerous and non-cancer tissues. However, when compared with tumor tissue, the difference in methylation between repeating elements and non-repeating regions in plasma DNA from non-cancer tissues and healthy individuals is even greater.

As a result, the methylation density of plasma DNA in cancer patients was significantly lower in repeating elements than in non-repeat regions. For repeating elements and non-repeat regions, the differences in plasma DNA methylation density between the mean values of four healthy controls and HCC patients were 0.163 and 0.088, respectively. Data from pre- and post-operative plasma samples also showed a larger dynamic range of methylation density changes in repeating regions than in non-repeat regions. In one embodiment, the plasma DNA methylation density of repeating elements can be used to determine whether a patient has cancer or to monitor disease progression.

As discussed above, variations in plasma methylation density in reference individuals also affect the accuracy of distinguishing between cancer patients and non-cancer individuals. A denser distribution of methylation density (i.e., a smaller standard deviation) results in more accurate differentiation between cancer and non-cancer individuals. In another embodiment, the coefficient of variation (CV) of methylation density in a 1 Mb region can be used as a criterion for selecting regions with low variability in plasma DNA methylation density within the reference population. For example, only regions <1% are selected. Other values, such as 0.5%, 0.75%, 1.25%, and 1.5%, can also be used as criteria for selecting regions with low variability in methylation density. In yet another embodiment, selection criteria may include both the CV of the region and the difference in methylation density between cancerous and non-cancer tissues.

When the methylation density of tumor tissue is known, it can also be used to assess the percentage concentration of tumor-derived DNA in plasma samples. This information can be obtained by analyzing a patient's tumor or by studying tumors from a large number of patients with the same type of cancer. As discussed above, plasma methylation density (P) can be expressed using the following equation: P = BKG × (1 - f) + TUM × f, where BKG is the background methylation density from blood cells and other organs, TUM is the methylation density in tumor tissue, and f is the percentage concentration of tumor-derived DNA in the plasma sample. This can be rewritten as:

BKG values can be determined by analyzing a patient's plasma sample at a time when cancer is absent, or by a study of a reference population of individuals without cancer. Therefore, f can be determined after measuring plasma methylation density.

F. Combination with other methods

The methylation analysis method described in this article can be used in combination with other methods based on genetic alterations in tumor-derived DNA in plasma. Examples of such methods include the analysis of cancer-related chromosomal abnormalities (Chen et al., 2013 Clinical Chemistry; 59:211-224; Liri et al., 2012 Science Translational Medicine; 4:162ra154) and cancer-related mononucleotide variants in plasma (Chen et al., 2013 Clinical Chemistry; 59:211-224). Methylation analysis methods have advantages over those genetic methods.

As shown in Figure 21A, hypomethylation of tumor DNA is a global phenomenon involving regions distributed across almost the entire genome. Therefore, DNA fragments from all chromosomal regions will be informative of the potential contribution of tumor-derived hypomethylated DNA to plasma/serum DNA in patients. In contrast, chromosomal abnormalities (amplification or deletion of chromosomal regions) are present only in some chromosomal regions, and DNA fragments from regions of tumor tissue without chromosomal abnormalities will not be informative in the analysis (Chen et al. 2013 Clinical Chemistry; 59: 211-224). Similarly, only a few thousand single nucleotide alterations are observed in each cancer genome (Chen et al. 2013 Clinical Chemistry; 59: 211-224). DNA fragments that do not overlap with these single nucleotide alterations will not be informative for determining the presence of tumor-derived DNA in plasma. Therefore, this methylation analysis method may be more cost-effective than those genetic methods used to detect cancer-related alterations in peripheral blood circulation.

In one embodiment, the cost-effectiveness of plasma DNA methylation analysis can be further enhanced by enriching DNA fragments from the most informative regions, such as those with the highest methylation differences between cancer and non-cancer tissues. Examples of methods for enriching these regions include using hybridization probes (such as the Nimblegen SeqCap system and the Agilent SureSelect target enrichment system), PCR amplification, and solid-phase hybridization.

G. Organize specific analysis/donors

Tumor-derived cells invade and metastasize to adjacent or distant organs. Invaded tissues or metastatic lesions contribute DNA to the plasma due to cell death. By analyzing the methylation patterns of DNA in the plasma of cancer patients and detecting the presence of tissue-specific methylation markers, tissue types associated with the disease process can be identified. This method provides a non-invasive anatomical scan of tissues associated with the cancer process to help identify primary and metastatic sites. Monitoring the relative concentrations of methylation markers in the plasma of the involved organs will also allow for assessment of the tumor burden in those organs and determination of whether the cancer process in those organs has regressed, improved, or has been cured. For example, if gene X is specifically methylated in the liver, then an increased concentration of methylated sequences from gene X in the plasma would be expected in cases of liver metastases associated with cancer (e.g., colorectal cancer). Another sequence or sequence group with similar methylation characteristics to gene X would also be present. Results from such sequences can then be combined. Similar considerations apply to other tissues, such as the brain, bone, lungs, and kidneys.

On the other hand, it is known that DNA from different organs exhibits tissue-specific methylation markers (Fotscher 2002 Nature Genetics; 31:175-179; BW Futscher et al. 2002 Nat Genet; 31:175-179; Zhan et al. 2008 Clinical Chemistry; 54:500-511). Therefore, analysis of methylation patterns in plasma can be used to elucidate the contribution of tissues from various organs to plasma. Elucidation of such contributions can be used to assess organ damage, as plasma DNA is believed to be released upon cell death. For example, liver lesions such as hepatitis (e.g., through viruses, autoimmune processes, etc.) or drug-induced liver toxicity (e.g., drug overdose (e.g., through paracetamol) or toxins (e.g., alcohols)) are associated with hepatocellular damage and will be expected to be associated with increased levels of liver-derived DNA in plasma. For example, if gene X is specifically methylated in the liver, then increased liver lesions will be expected to increase the concentration of methylated sequences from gene X in plasma. Conversely, if gene Y is particularly hypomethylated in the liver, then liver lesions will be expected to reduce the concentration of methylated sequences from gene Y in the plasma. In other embodiments, gene X or Y can be replaced by any genomic sequence that is not a gene and shows differences in methylation across different tissues in the body.

The techniques described in this article can also be applied to assess donor-derived DNA in the plasma of organ transplant recipients (Lo et al., 1998, The Lancet; 351:1329-1330). Polymorphic differences between donors and recipients have been used to distinguish donor-derived DNA from recipient-derived DNA in plasma (Zheng et al., 2012, Clinical Chemistry; 58:549-558). It has been proposed that tissue-specific methylation markers of transplanted organs can also be used as a method for detecting donor DNA in recipient plasma.

By monitoring the concentration of donor DNA, the condition of transplanted organs can be non-invasively assessed. For example, transplant rejection is associated with a high rate of cell death, and therefore, as reflected in the methylation characteristics of the transplanted organ, the concentration of donor DNA in the recipient's plasma (or serum) will be increased compared to patients in a stable condition or to other stable transplant recipients or healthy controls without transplantation. Similar to what has been described for cancer, donor-derived DNA can be identified in the recipient's plasma by detecting all or some of the characteristic features, including polymorphism, the shorter DNA size of the transplanted solid organ (Zheng et al. 2012 Clinical Chemistry; 58:549-558), and tissue-specific methylation patterns.

H. Size-normalized methylation

As described above by Lun et al. (Lun et al., Clinical Chemistry, 2013; doi:10.1373/clinchem.2013.212274), methylation density (e.g., plasma DNA) is related to the size of DNA fragments. The methylation density distribution of shorter plasma DNA fragments is significantly lower than that of longer fragments. It has been suggested that some non-cancerous conditions with anomalous plasma DNA fragmentation patterns (e.g., systemic lupus erythematosus (SLE)) may exhibit significant hypomethylation of plasma DNA due to the presence of a larger amount of shorter, less methylated plasma DNA fragments. In other words, the size distribution of plasma DNA may be a confounding factor for the methylation density of plasma DNA.

Figure 34A shows the size distribution of plasma DNA in SLE04 patients with SLE. The size distribution of the nine healthy controls is shown as gray dotted lines, and the size distribution of SLE04 is shown as a black solid line. Short plasma DNA fragments are more abundant in SLE04 than in the nine healthy controls. Because shorter DNA fragments are generally less methylated, this size distribution pattern may interfere with plasma DNA methylation analysis and cause more pronounced hypomethylation.

In some embodiments, the measured methylation levels can be normalized to reduce the interference of size distribution on plasma DNA methylation analysis. For example, the size of DNA molecules at multiple sites can be measured. In various implementations, the measurements can provide a specific size (e.g., length) of the DNA molecule, or simply determine that the size is within a specific range, which may also correspond to size. The normalized methylation level can then be compared to a threshold. Several methods exist for performing normalization to reduce the interference of size distribution on plasma DNA methylation analysis.

In one embodiment, size grading of DNA (e.g., plasma DNA) can be performed. Size grading ensures that DNA fragments of similar size are used in a manner consistent with a threshold for determining methylation levels. As part of size grading, DNA fragments with a first size (e.g., a first length range) can be selected, where a first threshold corresponds to the first size. Normalization can be achieved by calculating the methylation level using only the selected DNA fragments.

Size-grade separation can be achieved in various ways, such as through physical separation of DNA molecules of different sizes (e.g., by electrophoresis, or microfluidics-based techniques, or centrifugation-based techniques) or through computer simulation analysis. For computer simulation analysis, in one embodiment, massively parallel sequencing of the paired ends of plasma DNA molecules can be performed. The size of the sequenced molecule can then be inferred by comparing its position with that of each of the two ends of a plasma DNA molecule from a reference human genome. Subsequently, further analysis can be performed by selecting sequenced DNA molecules that match one or more size selection criteria (e.g., a size within a specified range). Thus, in one embodiment, the methylation density of fragments with similar sizes (e.g., within a specified range) can be analyzed. Thresholds (e.g., in box 2840 of method 2800) can be determined based on fragments within the same size range. For example, methylation levels can be determined from samples known to have or not have cancer, and thresholds can be determined from these methylation levels.

In another embodiment, a functional relationship between the methylation density and size of peripheral blood circulating DNA can be determined. This functional relationship can be defined by data points or function coefficients. The functional relationship can provide a correction value corresponding to the respective size (e.g., a shorter size may have a corresponding increase in methylation). In various implementations, the correction value can be between 0 and 1 or greater than 1.

Normalization can be based on average size. For example, the average size corresponding to the DNA molecule for calculating the first methylation level can be calculated, and the first methylation level can be multiplied by a corresponding correction value (i.e., corresponding to the average size). As another example, the methylation density of each DNA molecule can be normalized based on the size of the DNA molecule and the relationship between DNA size and methylation.

In another embodiment, normalization can be performed on a per-molecule basis. For example, the corresponding size of the DNA molecule at a specific site can be obtained (e.g., as described above), and the correction value corresponding to the corresponding size can be determined from a functional relationship. For unnormalized calculations, each molecule will be counted equally in terms of the methylation index at a given site. For normalized calculations, the contribution of a molecule to the methylation index can be weighted by a correction factor corresponding to the molecule size.

Figures 34B and 34C show the methylation analysis of plasma DNA from SLE patient SLE04 (Figure 34B) and HCC patient TBR36 (Figure 34C). The outer circle shows the Z- _methylation results of plasma DNA without computer-simulated size grading. The inner circle shows the Z- _methylation results of plasma DNA 130 bp or longer. For SLE patient SLE04, 84% of the region showed hypomethylation without computer-simulated size grading. When only fragments of 130 bp or longer were analyzed, the percentage of regions showing hypomethylation decreased to 15%. For HCC patient TBR36, 98.5% and 98.6% of the region showed plasma DNA hypomethylation with and without computer-simulated size grading, respectively. These results indicate that computer-simulated size grading can effectively reduce false-positive hypomethylation results associated with increased plasma DNA fragmentation in patients with conditions such as systemic lupus erythematosus or other inflammatory conditions.

In one embodiment, the results of analyses performed and not performed size grading can be compared to indicate whether size has any interfering effect on methylation results. Therefore, in addition to or instead of standardization, calculating the methylation level at a specific size can be used to determine whether the percentage of regions exceeding a threshold under size grading and without methylation differs, and whether there is a possibility of false positives, or whether only the specific methylation level differs. For example, the presence of a significant difference between the results of samples under size grading and without methylation can be used to indicate possible false positives due to abnormal fragmentation patterns. The threshold used to determine whether a difference is significant can be established by analyzing a cohort of cancer patients and a cohort of non-cancer controls.

I. Analysis of genome-wide CpG island hypermethylation in plasma

In addition to overall hypomethylation, hypermethylation of CpG islands is frequently observed in cancer (SB Baylin et al. 2011 Nat Rev Cancer; 11:726-734); Jones et al. 2007 Cell; 128:683-692). (M Esteller et al. 2007 Nat Rev Genet 2007; 8:286-298); M Ehrlich et al. 2002 Oncogene 2002; 21:5400-5413). In this section, genome-wide analysis using CpG island hypermethylation is described for the detection and monitoring of cancer.

Figure 35 is a flowchart of method 3500, which, according to an embodiment of the present invention, determines cancer grade classification based on the hypermethylation of CpG islands. In method 2800, the multiple sites may include CpG sites, wherein the CpG sites are organized into multiple CpG islands, each CpG island including one or more CpG sites. The methylation level of each CpG island can be used to determine the cancer grade classification.

At box 3510, the CpG islands to be analyzed are identified. In this analysis, as an example, a set of CpG islands to be analyzed is first identified, characterized by relatively low methylation density in the plasma of healthy reference individuals. In one aspect, the variation in methylation density in the reference population can be relatively small to allow for easier detection of cancer-related hypermethylation. In one embodiment, the CpG islands in the reference population have an average methylation density below a first percentage, and the coefficient of variation of methylation density in the reference population is below a second percentage.

As an example, to illustrate, the following criteria are used to identify applicable CpG islands:

i. The average methylation density of CpG islands in the reference population (e.g., healthy individuals) is <5%.

ii. The coefficient of variation for methylation density in plasma used to analyze a reference population (e.g., healthy individuals) is <30%.

These parameters can be adjusted for specific applications. From the dataset, 454 CpG islands in the genome meet these criteria.

At box 3520, the methylation density of each CpG island is calculated. The methylation density can be determined as described herein.

At box 3530, determine whether each CpG island is hypermethylated. For example, to analyze the hypermethylation of CpG islands in the test case, the methylation density of each CpG island is compared with the corresponding data in the reference population. The methylation density (an instance of methylation level) can be compared with one or more thresholds to determine whether a particular island is hypermethylated.

In one embodiment, a first threshold may correspond to the average methylation density of the reference population plus a specified percentage. Another threshold may correspond to the average methylation density of the reference population plus a specified number of standard deviations. In one implementation, a z-score (Z _-methylation ) is calculated and compared to the threshold. As an example, CpG islands in a test individual (e.g., individuals screening for cancer) are considered significantly hypermethylated if they meet the following criteria:

i. Its methylation density was higher than the reference population average by 2%, and

ii. Z _-methylation > 3.

These parameters can also be adjusted for specific applications.

At box 3540, the methylation density (e.g., z-score) of the hypermethylated CpG islands is used to determine the cumulative score. For example, after identifying all significantly hypermethylated CpG islands, a score involving the sum of z-scores or a function of the z-scores of all hypermethylated CpG islands can be calculated. One example of a score is the cumulative probability (CP) score, as described in another section. The cumulative probability score uses Z _-methylation to determine the probability of having such an observation by chance according to a probability distribution (e.g., Student's t-probability distribution with 3 degrees of freedom).

At box 3550, the cumulative score is compared to a cumulative threshold to determine the cancer grade classification. For example, if the total hypermethylation in the identified CpG islands is sufficiently high, the organism can be identified as having cancer. In one embodiment, the cumulative threshold corresponds to the highest cumulative score from the reference population.

IX. Methylation and CNA

As mentioned above, the methylation analysis method described herein can be used in combination with other methods based on genetic alterations in tumor-derived DNA in plasma. Examples of such methods include the analysis of cancer-related chromosomal abnormalities (Chen et al. 2013 Clinical Chemistry; 59:211-224; Leary et al. 2012 Science Translational Medicine; 4:162ra154). Aspects of copy number abnormalities (CNAs) are described in U.S. Patent Application No. 13/308,473.

A.CNA

Copy number abnormalities can be detected by counting DNA fragments aligned to a specific part of the genome, normalizing the counts, and comparing the counts to a threshold. In various embodiments, normalization can be performed by counting DNA fragments aligned to another haplotype of the same part of the genome (relative haplotype dose (RHDO)) or by counting DNA fragments aligned to another part of the genome.

The RHDO method relies on the use of heterozygous loci. The embodiments described in this section can also be used for homozygous loci, and are therefore non-haplotype-specific, by comparing two haplotypes of two regions rather than the same region. In the relative chromosomal region dosing method, the number of fragments from a chromosomal region (e.g., determined by counting sequence reads aligned to said region) is compared to a target value (which may come from a reference chromosomal region or from the same region in another known healthy sample). In this way, fragments are counted against the chromosomal region regardless of which haplotype the sequencing tag comes from. Therefore, sequence reads without heterozygous loci can still be used. For comparison, one embodiment may normalize the tag count before comparison. Each region is defined by at least two loci (separated from each other), and fragments at these loci can be used to obtain a cumulative value for said region.

A normalized value for the sequencing reads (tags) of a specific region can be calculated by dividing the number of sequencing reads aligned to that region by the total number of sequencing reads aligned to the whole genome. This normalized tag count can be generated by a sample whose results are to be compared with those of another sample. For example, the normalized value can be a proportion (e.g., a percentage or fraction) of the sequencing reads expected from the specific region, as described above. In other embodiments, other methods for normalization are possible. For example, normalization can be achieved by dividing the count of a region by the count of a reference region (in the above case, the reference region is exactly the whole genome). This normalized tag count can then be compared against a threshold that can be determined from one or more reference samples that do not show cancer.

The standardized tag count of the test case is then compared with the standardized tag counts of one or more reference individuals, such as cancer-free individuals. In one embodiment, the comparison is made by calculating a z-score for a specific chromosomal region. The z-score can be calculated using the following equation: z-score = (standardized tag count of the case - mean) / SD, where the "mean" is the mean standardized tag count compared to a specific chromosomal region of the reference sample; and SD is the standard deviation of the number of standardized tag counts compared to a specific region of the reference sample. Therefore, the z-score is the number of standard deviations of the standardized tag counts of the chromosomal region of the test case from the mean standardized tag counts of the same chromosomal region of one or more reference individuals.

In test organisms with cancer, amplified chromosomal regions in tumor tissue will appear excessively in plasma DNA. This will result in a positive z-score. Conversely, missing chromosomal regions in tumor tissue will appear insufficiently in plasma DNA. This will result in a negative z-score. The magnitude of the z-score is determined by several factors.

One factor is the percentage concentration of tumor-derived DNA in the biological sample (e.g., plasma). The higher the percentage concentration of tumor-derived DNA in the sample (e.g., plasma), the greater the difference between the normalized tag count of the test case and the reference case. Therefore, the z-score will be higher.

Another factor is the variation in standardized label counts across one or more reference cases. When the same degree of chromosomal region presentation is present in the biological sample (e.g., plasma) of the test case, a smaller variation (i.e., a smaller standard deviation) in the standardized label counts of the reference population will produce a higher z-score. Similarly, when the same degree of chromosomal region presentation is insufficient in the biological sample (e.g., plasma) of the test case, a smaller standard deviation in the standardized label counts of the reference population will produce a larger negative z-score.

Another factor is the magnitude of chromosomal abnormalities in tumor tissue. The magnitude of a chromosomal abnormality refers to a change in the copy number (increase or loss) of a specific chromosomal region. The higher the copy number change in tumor tissue, the greater the degree to which a specific chromosomal region is present in plasma DNA, either excessively or insufficiently. For example, the loss of two copies of a chromosome, compared to the loss of one of two copies, will result in a larger insufficient chromosomal region in plasma DNA, and therefore a larger negative z-score. Typically, multiple chromosomal abnormalities are present in cancer. The chromosomal abnormalities in each type of cancer can further vary in their nature (i.e., amplification or deletion), their degree (single or multiple copy increases or losses), and their extent (depending on the size of the chromosomal length abnormality).

The accuracy of measuring standardized tag counts is affected by the number of molecules analyzed. It is anticipated that 15,000, 60,000, and 240,000 molecules, respectively, will need to be analyzed to detect chromosomal abnormalities with a copy change (increase or loss), at percentage concentrations of approximately 12.5%, 6.3%, and 3.2%. Further details regarding tag counts for detecting cancer in different chromosomal regions are described in U.S. Patent Publication No. 2009/0029377, entitled “Diagnosis of Fetal Chromosomal Aneuploidy Using Massive Parallel Genome Sequencing,” by Yullo et al., the entire contents of which are incorporated herein by reference for all purposes.

The embodiments may also use size analysis instead of tag counting. Size analysis may also be used instead of normalized tag counting. Size analysis may use various parameters as mentioned herein and in U.S. Patent Application No. 12/940,992. For example, Q or F values from the above may be used. Such size values do not need to be normalized by counting from other regions because these values are not proportionally adjusted with the number of readings. Haplotype-specific methodological techniques, such as the RHDO method described in more detail above and in U.S. Patent Application No. 13/308473, may also be used for non-specific methods. For example, techniques involving sequencing depth and region optimization may be used. In some embodiments, GC preference for a particular region may be considered when comparing two regions. Because the RHDO method uses the same regions, this type of correction is not required.

While some cancers may typically exhibit abnormalities in specific chromosomal regions, these cancers do not always present abnormalities only in these regions. For example, additional chromosomal regions may show abnormalities, and the location of such additional regions may be unknown. Furthermore, when screening patients to identify early-stage cancers, it may be desirable to identify a large number of various types of cancer that may show abnormalities at any location across the entire genome. To address these situations, embodiments can systematically analyze multiple regions to determine which regions show abnormalities. The number of abnormalities and their location (e.g., whether they are adjacent) can be used, for example, to confirm abnormalities, determine cancer stage, provide a cancer diagnosis (e.g., whether the number exceeds a threshold), and provide prognosis based on the number and location of the individual regions showing abnormalities.

Therefore, embodiments can identify whether an organism has cancer based on the number of regions displaying abnormalities. Thus, multiple regions (e.g., 3,000) can be tested to identify the number of regions displaying abnormalities. These regions can cover the entire genome or only portions of the genome, such as non-repetitive regions.

Figure 36 is a flowchart of method 3600, which, according to an embodiment of the present invention, analyzes biological samples of an organism using multiple chromosomal regions. The biological samples include nucleic acid molecules (also referred to as fragments).

At box 3610, multiple regions (e.g., non-overlapping regions) of an organism's genome are identified. Each chromosomal region includes multiple loci. The size of the region can be 1 Mb, or some other equivalent size. In the case of a 1 Mb region, the entire genome could then include approximately 3,000 regions, each with a predetermined size and location. Such predetermined regions can vary to accommodate a specific chromosome length or a specified number of regions to be used, and any other criteria mentioned herein. If regions have different lengths, such lengths can be used to standardize the results, for example, as described herein. Regions can be specifically selected based on certain criteria for a particular organism and/or based on knowledge of the cancer being tested. Regions can also be selected arbitrarily.

At box 3620, for each of the multiple nucleic acid molecules, the location of the nucleic acid molecule in the reference genome of the organism is identified. The location can be determined using any of the methods mentioned herein, such as obtaining a sequencing tag from a sequencing fragment and aligning the sequencing tag to the reference genome. A specific haplotype of the molecule can also be determined using haplotype-specific methods.

Boxes 3630-3650 are defined for each chromosomal region. At box 3630, nucleic acid molecules from a corresponding population are identified as originating from the chromosomal region based on their location. A corresponding population may include nucleic acid molecules located at at least one of multiple loci within the chromosomal region. In one embodiment, the population may be a fragment aligned to a specific haplotype of the chromosomal region, such as in the RHDO method described above. In another embodiment, the population may be any fragment aligned to a chromosomal region.

At box 3640, the computer system calculates the corresponding value for the nucleic acid molecules in the corresponding population. The corresponding value defines the characteristics of the nucleic acid molecules in the corresponding population. The corresponding value can be any value mentioned herein. For example, the value can be the number of fragments in the population or a statistical value of the size distribution of fragments in the population. The corresponding value can also be a standardized value. For example, the tag count of a region divided by the total number of tag counts in the sample or the number of tag counts in a reference region. The corresponding value can also be a difference or ratio to another value (e.g., in RHDO), thereby providing the differential characteristics of the region.

At box 3650, the corresponding value is compared to a reference value to determine whether the first chromosomal region shows a deletion or an amplification. This reference value can be any threshold or reference value described herein. For example, a reference value can be a threshold determined for a normal sample. For RHDO, the corresponding value can be the difference or ratio of the tag counts of two haplotypes, and the reference value can be a threshold used to determine if there is a statistically significant deviation. As another example, the reference value can be the tag count or size value of another haplotype or region, and the comparison can include taking the difference or ratio (or a function of them) and subsequently determining whether the difference or ratio exceeds a threshold.

The reference value can vary based on results from other regions. For example, if neighboring regions also show bias (but less than a threshold, such as a z-score of 3), then a lower threshold can be used. For instance, if three adjacent regions all exceed a first threshold, then cancer is more likely. Therefore, this first threshold can be lower than another threshold required to identify cancer from non-adjacent regions. Three regions (or more) with even smaller biases can have a sufficiently low probability corresponding to random fluctuations, thus maintaining sensitivity and specificity.

At box 3660, determine the number of genomic regions classified as showing deletions or amplifications. The chromosomal regions counted may be limited. For example, only regions adjacent to at least one other region may be counted (or adjacent regions may need to have a certain size, such as four or more regions). For embodiments where the regions are not identical, the number may also consider the corresponding length (e.g., the number could be the total length of the aberrant regions).

At box 3670, the quantity is compared to a quantity threshold to determine the classification of the sample. As an example, the classification could be whether an organism has cancer, the stage of cancer, and the prognosis of cancer. In one embodiment, all abnormal regions are counted and a single threshold is applied, regardless of where the region appears. In another embodiment, the threshold can vary based on the location and size of the counted regions. For example, the quantity of a region on a specific chromosome or chromosome arm can be compared to a threshold for that specific chromosome (or arm). Multiple thresholds can be used. For example, the quantity of abnormal regions on a specific chromosome (or arm) must exceed a first threshold, and the total number of abnormal regions in the genome must exceed a second threshold. The threshold can be a percentage of regions showing deletions or amplifications.

The threshold for the quantity of a region can also depend on how strong the imbalance is claimed among the regions being counted. For example, the quantity of a region used as a threshold to determine a cancer classification might depend on the specificity and sensitivity (abnormality threshold) used to detect abnormalities in each region. For instance, if the abnormality threshold is low (e.g., a z-score of 2), then a high quantity threshold (e.g., 150) can be chosen. But if the abnormality threshold is high (e.g., a z-score of 3), then the quantity threshold can be low (e.g., 50). The quantity of regions exhibiting abnormalities can also be a weighted value; for example, a region exhibiting high imbalance can be weighted higher than a region exhibiting only slight imbalance (i.e., more classifications than simply abnormal positive and negative). As an example, the sum of z-scores could be used, thus employing a weighted value.

Therefore, the amount of significantly overrepresented or underrepresented chromosomal regions (which may include number and/or size) exhibiting normalized tag counts (or other corresponding values for population characteristics) can be used to reflect disease severity. The amount of chromosomal regions with abnormal normalized tag counts can be determined by two factors: the number (or size) of chromosomal abnormalities in tumor tissue and the percentage concentration of tumor-derived DNA in a biological sample (e.g., plasma). More advanced cancers tend to exhibit more (and larger) chromosomal abnormalities. Therefore, more cancer-related chromosomal abnormalities are likely to be detectable in samples (e.g., plasma). In patients with more advanced cancer, a higher tumor burden will result in a higher percentage concentration of tumor-derived DNA in the plasma. As a result, tumor-related chromosomal abnormalities will be more easily detected in plasma samples.

One possible approach to improve sensitivity without compromising specificity is to consider the results of adjacent chromosomal segments. In one embodiment, the z-score threshold remains >2 and <-2. However, the chromosomal region is only classified as potentially abnormal if two consecutive segments are likely to show the same type of abnormality, for example, if the z-scores of both segments are >2. In other embodiments, the z-scores of adjacent segments can be summed using a higher threshold. For example, the z-scores of three consecutive segments can be summed and a threshold of 5 can be used. This concept can be extended to more than three consecutive segments.

The combination of quantity and abnormality threshold may also depend on the purpose of the analysis and any prior knowledge (or lack thereof) about the organism. For example, if screening a normal healthy population for cancer, then a high specificity would typically be used in terms of the quantity of regions (i.e., a high threshold for the number of regions) versus the abnormality threshold when a region is identified as having an abnormality. However, in patients at higher risk (e.g., those with a tumor or family history, smokers, chronic human papillomavirus (HPV) carriers, hepatitis virus carriers, or other viral carriers), the threshold may be lower for higher sensitivity (fewer false negatives).

In one embodiment, if chromosomal abnormalities are detected using a 1Mb resolution and a detection limit of 6.3% for tumor-derived DNA, then the number of molecules in each 1Mb segment would need to be 60,000. For the whole genome, this would translate to approximately 180 million (60,000 readings/Megabase × 3,000 Megabase) comparable readings.

Smaller segment sizes will produce higher resolution for detecting smaller chromosomal abnormalities. However, this will increase the total number of molecules required to be analyzed. Larger segment sizes, at the expense of resolution, will reduce the number of molecules required for analysis. Therefore, only larger abnormalities can be detected. In one implementation, a larger region can be used, and the segment showing the abnormality can be subdivided and analyzed to obtain better resolution (e.g., as described above). The number of molecules to be analyzed can be determined by evaluating the size of the deletion or amplification to be detected (or the minimum concentration to be detected).

B. CNA based on sequencing of bisulfite-treated plasma DNA

Genome-wide hypomethylation and CNA can frequently be observed in tumor tissue. Here, it is demonstrated that information on CNA and cancer-related methylation alterations can be obtained simultaneously from bisulfite sequencing of plasma DNA. Because both types of analysis can be performed on the same dataset, there is practically no additional cost for CNA analysis. Other embodiments can use different procedures to obtain methylation and genetic information. In other embodiments, similar analyses can be performed on cancer-related hypermethylation in conjunction with CNA analysis.

Figure 37A shows the CNA analysis of tumor tissue, untreated plasma DNA, and BS-treated plasma DNA (inside to outside) from patient TBR36. The outermost ring shows the chromosome G-banding diagram. Each point represents the result for a 1 Mb region. Green, red, and gray dots represent regions of copy number increase, copy number loss, and no copy number change, respectively. For plasma analysis, z-scores are shown. A difference of 5 exists between the two concentric lines. For tumor tissue analysis, copy number is shown. A copy difference of one exists between the two concentric lines. Figure 38A shows the CNA analysis of tumor tissue, untreated plasma DNA, and BS-treated plasma DNA (inside to outside) from patient TBR34. The CNA patterns detected in both bisulfite-treated and untreated plasma samples were consistent.

The patterns of CNAs detected in tumor tissue, untreated plasma, and bisulfite-treated plasma were consistent. To further evaluate the consistency between the results of bisulfite-treated and untreated plasma, scatter plots were constructed. Figure 37B is a scatter plot showing the relationship between z-scores of CNAs in the 1Mb region detected using bisulfite-treated plasma and untreated plasma for patient TBR36. A positive correlation was observed between the z-scores of the two analyses (r = 0.89, p < 0.001, Pearson correlation). Figure 38B is a scatter plot showing the relationship between z-scores of CNAs in the 1Mb region detected using bisulfite-treated plasma and untreated plasma for patient TBR34. A positive correlation was observed between the z-scores of the two analyses (r = 0.81, p < 0.001, Pearson correlation).

C. Synergistic analysis of cancer-related CNA and methylation alterations

As described above, CNA analysis can include counting the number of sequence reads in each 1Mb region, while methylation density analysis can include detecting the proportion of methylated cytosine residues on CpG dinucleotides. The combination of these two analyses can produce synergistic information for cancer detection. For example, methylation classification and CNA classification can be used to determine the third category of cancer severity.

In one embodiment, the presence of either cancer-related CNA or methylation alterations can be used to indicate the possible presence of cancer. In such embodiments, the sensitivity of cancer detection can be increased when both CNA and methylation alterations are present in the plasma of the test individual. In another embodiment, the presence of both alterations can be used to indicate the presence of cancer. In such embodiments, the specificity of the test can be improved because either type of alteration can be detected in some non-cancer individuals. Therefore, the third category is likely to be cancer-positive only if both the first and second categories indicate cancer.

Twenty-six HCC patients and 22 healthy individuals were recruited. Blood samples were collected from each individual, and plasma DNA was sequenced after bisulfite treatment. For HCC patients, blood samples were collected at diagnosis. The presence of significant amounts of CNA was defined, for example, as >5% of the region showing a z-score <-3 or >3. The presence of significant amounts of cancer-associated hypomethylation was defined as >3% of the region showing a z-score <-3. As examples, the amount of a region is expressed as the raw count, percentage, and length of the region.

Table 3 shows the significant CNA and methylation alterations in the plasma of 26 HCC patients detected by massively parallel sequencing of bisulfite-treated plasma DNA.

Table 3

The detection rates for cancer-related methylation alterations and CNA were 69% and 50%, respectively. If the presence of either criterion is used to indicate the possible presence of cancer, the detection rate (i.e., diagnostic sensitivity) increases to 73%.

Results are presented for two patients with either CNA (Figure 39A) or methylation alterations (Figure 39B). Figure 39A is a Circos diagram showing CNA (inner ring) and methylation (outer ring) analysis of bisulfite-treated plasma from HCC patient TBR240. For CNA analysis, green, red, and gray dots represent regions of increased, decreased, and no copy number alterations, respectively. For methylation analysis, green, red, and gray dots represent regions with high, low, and normal methylation, respectively. In this patient, cancer-associated CNA was detected in the plasma, but methylation analysis did not reveal a significant amount of cancer-associated low methylation. Figure 39B is a Circos diagram showing CNA (inner ring) and methylation (outer ring) analysis of bisulfite-treated plasma from HCC patient TBR164. In this patient, cancer-associated low methylation was detected in the plasma. However, a significant amount of CNA could not be observed. The results of two patients with CNA and methylation alterations are shown in Figures 48A (TBR36) and 49A (TBR34).

Table 4 shows the significant CNA and methylation alterations in the plasma of 22 control individuals detected using massively parallel sequencing on bisulfite-treated plasma DNA. Random sampling (i.e., leave-one-out crossover test) was used to evaluate each control individual. Therefore, when evaluating a particular individual, the other 21 individuals were used to calculate the mean and SD of the control group.

Table 4

The detection specificity for significant methylation alterations and CNA was 86% and 91%, respectively. If the presence of both criteria is required to indicate the possible presence of cancer, the specificity increases to 95%.

In one embodiment, a sample positive for both CNA and/or hypomethylation is considered cancer-positive, and a sample negative is considered negative when both are undetectable. Using "OR" logic provides higher sensitivity. In another embodiment, only samples positive for both CNA and hypomethylation are considered cancer-positive, thus providing higher specificity. In yet another embodiment, a three-tiered classification can be used. Individuals are classified as i. both normal; ii. one abnormal; iii. both abnormal.

Different follow-up strategies can be used for these three categories. For example, individuals in category (iii) may undergo the most intensive follow-up protocol, such as involving whole-body imaging; individuals in category (ii) may undergo a less intensive follow-up protocol, such as repeat plasma DNA sequencing at relatively short intervals of several weeks; and individuals in category (i) may undergo the least intensive follow-up protocol, such as retesting after many years. In other embodiments, methylation and CNA measurements may be used in conjunction with other clinical parameters, such as imaging results or serum biochemistry, to further optimize the classification.

D. Prognostic value of plasma DNA analysis after treatment

The presence of cancer-related CNA and/or methylation alterations in plasma indicates the presence of tumor-derived DNA in the circulation of cancer patients. These cancer-related alterations are expected to decrease or clear after treatment (e.g., surgery). Conversely, the persistence of these alterations in plasma after treatment can indicate that all tumor cells from the body have not been completely eliminated and can serve as a suitable predictor of disease recurrence.

One week after surgical resection of the tumor intended for cure, blood samples were collected from two HCC patients, TBR34 and TBR36. CNA and methylation analyses were performed on the post-treatment plasma samples after bisulfite treatment.

Figure 40A shows the CNA analysis of bisulfite-treated plasma DNA before (inner loop) and after (outer loop) tumor resection in HCC patient TBR36. Each dot represents the result for a 1 Mb region. Green, red, and gray dots represent regions of copy number increase, copy number loss, and no copy number change, respectively. Most of the CNA observed before treatment disappeared after tumor resection. The proportion of regions with z-scores <-3 or >3 decreased from 25% to 6.6%.

Figure 40B shows the methylation analysis of bisulfite-treated plasma DNA in HCC patient TBR36 before (inner ring) and after (outer ring) tumor resection. Green, red, and gray dots represent regions with hypermethylation, hypomethylation, and normal methylation, respectively. The proportion of regions showing significant hypomethylation decreased significantly from 90% to 7.9%, and the degree of hypomethylation also showed a significant reduction. This patient achieved complete clinical remission 22 months after tumor resection.

Figure 41A shows the CNA analysis of bisulfite-treated plasma DNA before (inner loop) and after (outer loop) tumor resection in HCC patient TBR34. Although the number and magnitude of CNA-displaying regions decreased in the affected areas after tumor resection, residual CNA was still observable in postoperative plasma samples. The red circles highlight the areas with the most prominent residual CNA. The proportion of regions displaying z-scores <-3 or >3 decreased from 57% to 12%.

Figure 41B shows the methylation analysis of bisulfite-treated plasma DNA before (inner loop) and after (outer loop) tumor resection in HCC patient TBR34. The amount of hypomethylation decreased after tumor resection, with the mean z-score of hypomethylated regions decreasing from -7.9 to -4.0. However, the proportion of regions with z-scores <-3 showed the opposite change, increasing from 41% to 85%. This observation may indicate the presence of residual cancer cells after treatment. Clinically, multiple lesions with tumor nodules were detected in the remaining unresected liver 3 months after tumor resection. Lung cancer metastasis was observed 4 months post-surgery. The patient died 8 months post-surgery from local recurrence and metastatic disease.

The observations in these two patients (TBR34 and TBR36) suggest that the presence of residual cancer-related alterations in CNA and hypomethylation can be used to monitor and predict cancer patients after treatment intended for cure. The data also demonstrate that the degree of alteration in the amount of plasma CNA detected can be used in conjunction with the assessment of alterations in plasma DNA hypomethylation to predict and monitor treatment efficacy.

Therefore, in some embodiments, a biological sample is obtained before treatment and a second biological sample is obtained after treatment (e.g., surgery). A first value is obtained for the first sample, such as a z-score (e.g., a normalized value of the region's methylation level and CNA) and the number of regions exhibiting hypomethylation and CNA (e.g., amplification or deletion). A second value can be obtained for the second sample. In another embodiment, a third or even additional sample can be obtained after treatment. The number of regions exhibiting hypomethylation and CNA (e.g., amplification or deletion) can be obtained from the third or even additional sample.

As described above with respect to Figures 40A and 41A, a first number of hypomethylated regions in the first sample can be compared with a second amount of hypomethylated regions in the second sample. As described above with respect to Figures 40B and 41B, a first amount of hypomethylated regions in the first sample can be compared with a second amount of hypomethylated regions in the second sample. Comparisons of the first amount with the second amount and the first number with the second number can be used to determine the prognosis of treatment. In different embodiments, only one comparison may be used to determine the prognosis, or both comparisons may be used. In embodiments where a third or even additional samples are obtained, one or more of these samples may be used alone or in combination with the second sample to determine the prognosis of treatment.

In one implementation, a worse prognosis is predicted when the first difference between the first and second amounts is below a first difference threshold. In another embodiment, a worse prognosis is predicted when the second difference between the first and second amounts is below a second difference threshold. The thresholds can be the same or different. In one embodiment, the first and second difference thresholds are zero. Therefore, for the above examples, the difference between methylation values will indicate a worse prognosis for the patient with TBR34.

If the first and/or second differences exceed the same or corresponding thresholds, the prognosis can be better. The classification of the prognosis can depend on how much the differences are below or above the thresholds. Multiple thresholds can be used to provide a variety of classifications. Larger differences predict better outcomes, while smaller differences (and even negative values) predict worse outcomes.

In some embodiments, the time points at which each sample was collected are also recorded. Under such time parameters, the rate of change of kinetics or quantities can be determined. In one embodiment, a rapid decrease in plasma tumor-associated hypomethylation and/or a rapid decrease in plasma tumor-associated CNA predicts a favorable prognosis. Conversely, a static or rapid increase in plasma tumor-associated hypomethylation and/or a static or rapid increase in plasma tumor-associated CNA predicts a poor prognosis. Methylation and CNA measurements can be used in conjunction with other clinical parameters (e.g., imaging findings or serum biochemical or protein markers) to predict clinical outcomes.

In addition to plasma, other samples may be used in the examples. For example, tumor-related methylation abnormalities (e.g., hypomethylation) and/or tumor-related CNAs can be measured from circulating tumor cells in the blood of cancer patients, or from cell-free DNA or tumor cells in urine, feces, saliva, sputum, bile, pancreatic fluid, cervical swabs, genital tract (e.g., vaginal) secretions, ascites, pleural fluid, semen, sweat, and tears.

In various embodiments, tumor-related methylation abnormalities (e.g., hypomethylation) and/or tumor-related CNAs can be detected from the blood or plasma of patients with breast cancer, lung cancer, colorectal cancer, pancreatic cancer, ovarian cancer, nasopharyngeal carcinoma, cervical cancer, melanoma, brain tumors, etc. In fact, because methylation and genetic alterations such as CNA are common in cancer, the methods described can be used for all cancer types. Methylation and CNA measurements can be used in conjunction with other clinical parameters (e.g., imaging results) to predict clinical outcomes. The embodiments can also be used to screen and monitor patients with pre-tumor lesions, such as adenomas.

Therefore, in one embodiment, the biological sample is collected prior to treatment, and CNA and methylation measurements are repeated post-treatment. The measurements can yield a subsequent first quantity identifying regions showing deletion or amplification, and a subsequent second quantity identifying regions where the methylation level exceeds a corresponding regional threshold. The first quantity can be compared with the subsequent first quantity, and the second quantity can be compared with the subsequent second quantity to determine the prognosis of the organism.

Comparisons used to determine the prognosis of an organism may include determining a first difference between a first quantity and a subsequent first quantity, and this first difference may be compared to one or more first difference thresholds to determine the prognosis. Comparisons used to determine the prognosis of an organism may also include determining a second difference between a second quantity and a subsequent second quantity, and this second difference may be compared to one or more second difference thresholds. The thresholds may be zero or another number.

The prognosis can be predicted to be worse when the first difference is below a first difference threshold than when the first difference exceeds a first difference threshold. Similarly, the prognosis can be predicted to be worse when the second difference is below a second difference threshold than when the second difference exceeds a second difference threshold. Examples of treatments include immunotherapy, surgery, radiation therapy, chemotherapy, antibody-based therapies, gene therapy, epigenetic therapy, or targeted therapy.

E. Performance

The diagnostic performance for CNA and methylation analysis with different numbers of sequence reads and region sizes is described below.

1. Number of sequence readings

According to one embodiment, plasma DNA was analyzed from 32 healthy controls, 26 patients with hepatocellular carcinoma, and 20 patients with other types of cancer, including nasopharyngeal carcinoma, breast cancer, lung cancer, neuroendocrine cancer, and leiomyosarcoma. Twenty-two of the 32 healthy individuals were randomly selected as a reference group. The mean and standard deviation (SD) of these 22 reference individuals were used to determine the normal range for methylation density and genome representation. DNA extracted from plasma samples from each individual was used to construct sequencing libraries using an Illumina paired-end sequencing kit. The sequencing libraries were then subjected to bisulfite treatment to convert unmethylated cytosine residues into uracil. The bisulfite-converted sequencing libraries from each plasma sample were sequenced using one channel of an Illumina HiSeq2000 sequencer.

After base determination, linker sequences and low-quality bases (i.e., quality fraction <5) at the ends of fragments were removed. The pruned reads in FASTQ format were then processed using a bioinformatics workflow called Methy-Pipe for methylation data analysis (Jiang et al. 2010, IEEE International Conference on Bioinformatics and Biomedicine, doi:10.1109/BIBMW.2010.5703866). To align the bisulfite-converted sequencing reads, they were first aligned with a reference human genome (NCBI build 36/hg19), and a computer program was used to separately convert all cytosine residues to thymine in the Watson and Crick chains. Subsequently, all processed reads underwent a cytosine-to-thymine transformation, and the positional information of each transformed residue was preserved. The transformed reads were aligned using SOAP2 with two transformed reference human genomes (R Li et al., 2009 Bioinformatics 25:1966-1967), with a maximum of two mismatches allowed per aligned read. Only reads aligned to unique genomic locations were used for downstream analysis. Indistinct reads and duplicate (clonal) reads aligned to both Watson and Crick chains were removed. Cytosine residues in the CpG dinucleotide background were used for downstream methylation analysis. After alignment, cytosine originally present on the sequencing reads was recovered based on the positional information preserved during the computer-programmed transformation. The recovered cytosine in CpG dinucleotides was scored as methylated. The thymine in CpG dinucleotides was scored as unmethylated.

For methylation analysis, the genome is divided into regions of uniform size. Regions tested include sizes of 50kb, 100kb, 200kb, and 1Mb. The methylation density of each region is calculated as the number of methylated cytosines against a CpG dinucleotide background divided by the total number of cytosines at CpG sites. In other embodiments, the region sizes may vary across the genome. In one embodiment, each of these regions of varying sizes is compared across multiple individuals.

To determine whether the plasma methylation density of the test cases was normal, the methylation density was compared with the results of a reference group. Twenty-two of the 32 randomly selected healthy individuals were used as the reference group to calculate the methylation z-score (Z _-methylation ).

Wherein, MD _test is the methylation density of test cases in a specific 1Mb region; is the average methylation density of the reference population in the corresponding region; and MD _SD is the SD of the methylation density of the reference population in the corresponding region.

For CNA analysis, the number of sequencing reads aligned to each 1Mb region was determined (Chen et al. 2013 Clinical Chemistry 59:211-24). The sequencing read density for each region, corrected for GC preference, was determined using locally weighted regression scatter smoothing as previously described (Chen et al. 2011 PLoS One 6:e21791). For plasma analysis, the sequencing read density of the test cases was compared to that of the reference population to calculate the z-score of the CNA (Z _CNA ).

The RD _test is the sequencing read density of a test case in a specific 1Mb region; the average sequencing read density of the reference population for the corresponding region; and the RD _SD is the SD of the sequencing read density of the reference population for the corresponding region. If the Z _CNA of a region is <-3 or >3, then the region is defined by a CNA.

Each case yielded an average of 93 million aligned reads (range: 39 million to 142 million). To assess the impact of reduced sequencing read counts on diagnostic performance, 10 million aligned reads were randomly selected from each case. Reference individuals from the same group were used to establish reference ranges for each 1Mb region in the dataset with reduced sequencing read counts. The percentage of regions exhibiting significant hypomethylation (Z _methylation < -3) and having CNA (Z _CNA < -3 or > 3) were determined for each case. Receiver operating characteristic (ROC) curves were used to illustrate the diagnostic performance of whole-genome hypomethylation and CNA analysis for datasets with all sequencing reads from one channel and 10 million reads per case. All 32 healthy individuals were used in the ROC analysis.

Figure 42 illustrates the diagnostic performance of whole-genome hypomethylation analysis with varying numbers of sequencing reads. For hypomethylation analysis, the area under the ROC curve did not differ significantly between the two datasets, which analyzed all sequencing reads from one channel and 10 million reads per case (P = 0.761). For CNA analysis, diagnostic performance degraded significantly with a decrease in the area under the curve as the number of sequencing reads decreased from 10 million using data from one channel (P < 0.001).

2. The impact of using different area sizes

In addition to dividing the genome into 1Mb regions, the possibility of using smaller region sizes was explored. Theoretically, using smaller regions could reduce the variability in methylation density within those regions. This is because methylation density can vary significantly between different genomic regions. Larger regions increase the likelihood of including regions with different methylation densities, thus leading to an overall increase in the variability in methylation density within those regions.

While using smaller region sizes may reduce variability in methylation density related to inter-regional differences, it also reduces the number of sequencing reads aligned to a specific region. Reduced reads aligned to individual regions increase variability due to sampling variations. The optimal region size that minimizes overall variability in methylation density can be determined experimentally for the needs of a specific diagnostic application, such as the total number of sequencing reads per sample and the type of DNA sequencer used.

Figure 43 is a graph showing the ROC curves for detecting cancer based on whole-genome hypomethylation analysis using different region sizes (50kb, 100kb, 200kb, and 1Mb). The p-values shown are comparisons of the area under the curve for the 1Mb region size. A tendency for improvement can be seen when the region size is reduced from 1Mb to 200kb.

F. Cumulative probability score

The amount of methylated and CNA regions can be various values. The examples above describe the number of regions exceeding a threshold or the percentage of such regions exhibiting significant hypomethylation or CNA as a parameter for classifying whether a sample is associated with cancer. Such methods do not account for anomalous values in individual regions. For example, a region with Z _{-methylation of} -3.5 will be considered the same as a region with Z _-methylation of -30, as both will be classified as having significant hypomethylation. However, the degree of hypomethylation alteration in plasma, i.e., the magnitude of Z _-methylation values, is influenced by the amount of cancer-associated DNA in the sample, and therefore, the percentage of regions exhibiting anomalous values can be used to supplement information reflecting tumor burden. A higher percentage concentration of tumor DNA in a plasma sample will result in a lower hypomethylation density, which will translate into a lower Z- _methylation value.

1. Cumulative probability score as a diagnostic parameter

To utilize information from outlier values, a method called cumulative probability (CP) score was developed. Based on a normal distribution probability function, each Z _-methylation value is transformed into the probability of having such an observation by chance.

The CP score is calculated as follows:

For regions (i) with Z _-methylation < -3, the CP score is ∑-log(Prob _i ).

Where Prob _{_i} is the probability of Z _-methylation of region (i) according to a Steudent t-distribution with 3 degrees of freedom, and log_i is the natural logarithm function. In another embodiment, a logarithm with a base of 10 (or other number) can be used. In other embodiments, other distributions, such as (but not limited to) normal and gamma distributions, can be used to transform the z-scores into CP.

A higher CP score indicates a lower probability that a normal population happens to have this type of deviated methylation density. Therefore, a high CP score will indicate a higher probability of having abnormally low methylation DNA in the sample, such as cancer-related DNA.

Compared to the percentage of areas showing abnormalities, CP score measurements have a wider dynamic range. While tumor burden can vary considerably between patients, a wider range of CP values will be applicable to reflecting the tumor burden in patients with relatively high and relatively low tumor burdens. Furthermore, the use of CP scores may be more sensitive for detecting changes in the concentration of tumor-associated DNA in plasma. This is advantageous for monitoring treatment response and prediction. Therefore, a decrease in CP score during treatment indicates a good response to treatment. The lack of a decrease in CP score or even an increase during treatment will indicate a poor or absent response. For prognosis, a high CP score indicates a high tumor burden and suggests a poor prognosis (e.g., a higher chance of death or tumor progression).

Figure 44A illustrates the diagnostic performance of cumulative probability (CP) and the percentage of regions with anomalies. There was no significant difference in the area under the curve between the two types of diagnostic algorithms (P = 0.791).

Figure 44B illustrates the diagnostic performance of plasma analyses for overall hypomethylation, CpG island hypermethylation, and CNA. With one channel sequenced per sample (hypomethylation analysis was defined as a 200 kb region size, CNA as a 1 Mb region size, and CpG islands as defined by the database hosted by the University of California, Santa Cruz (UCSC), the area under the curve (AUC) for all three analysis types was above 0.90.

In subsequent analyses, the highest CP score in the control individuals was used as a threshold for each of the three types of analyses. These thresholds were selected to achieve 100% diagnostic specificity. The diagnostic sensitivities for overall hypomethylation, CpG island hypermethylation, and CNA analyses were 78%, 89%, and 52%, respectively. At least one of the three types of abnormalities was detected in 43 out of 46 cancer patients, thus yielding a sensitivity of 93.4% and a specificity of 100%. These results indicate that the three types of analyses can be used synergistically to detect cancer.

Figure 45 shows the results for overall hypomethylation, CpG island hypermethylation, and CNA in patients with hepatocellular carcinoma. The CP score thresholds for the three types of analysis are 960, 2.9, and 211, respectively. Positive CP score results are shown in bold and underlined.

Figure 46 shows the results for overall hypomethylation, CpG island hypermethylation, and CNA in patients with cancers other than hepatocellular carcinoma. The CP score thresholds for the three types of analysis are 960, 2.9, and 211, respectively. Positive CP score results are shown in bold and underlined.

2. Application of CP score in cancer surveillance

Series of samples were collected from TBR34 of HCC patients before and after treatment. Overall hypomethylation of the samples was analyzed.

Figure 47 shows a series of analyses of plasma methylation in case TBR34. The innermost ring shows the methylation density of the leukocyte layer (black) and tumor tissue (purple). For plasma samples, Z- _methylation is shown for each 1Mb region. The difference between the two lines represents a 5-fold difference in Z _-methylation . Red and gray dots represent regions with low methylation compared to the reference population and where the methylation density remained unchanged. From the second innermost ring outwards, plasma samples were collected before treatment, 3 days after tumor resection, and 2 months later, respectively. Before treatment, high levels of low methylation were observed in the plasma, with over 18.5% of regions having Z- _methylation <-10. 3 days after tumor resection, a decrease in low methylation was observed in the plasma, with no regions having Z _-methylation <-10.

Table 5

Table 5 shows that although the magnitude of hypomethylation changes decreased 3 days after tumor resection, the percentage of areas showing abnormalities showed an abnormal increase. On the other hand, CP scores more accurately reveal the decrease in plasma hypomethylation and can better reflect changes in tumor burden.

Two months after surgery, a significant percentage of areas exhibiting hypomethylation remained. The CP score also remained fixed at approximately 15,000. This patient was later diagnosed with multifocal tumors (previously unknown at surgery) in the remaining unresected liver three months after surgery, and multiple lung cancer metastases were noted four months after surgery. The patient died of metastatic disease eight months after surgery. These results suggest that the CP score may be a more effective indicator of tumor burden than the percentage of areas with abnormalities.

In general, CP can be applied to applications that require measuring the amount of tumor DNA in plasma. Examples of such applications include: prognostic and monitoring cancer patients (e.g., observing response to treatment or monitoring tumor progression).

The cumulative z-score is a direct sum of z-scores, meaning it is not converted into a probability. In this example, the cumulative z-score exhibits the same behavior as the CP score. In other cases, the CP may be more sensitive than the cumulative z-score in monitoring residual disease because the CP score has a larger dynamic range.

The effect of X.CNA on methylation

The preceding text describes the use of CNA and methylation in the corresponding classification for determining cancer grade, where classifications are combined to provide a third category. In addition to this combination, CNA can also be used to modify the threshold of methylation analysis and to identify false positives by comparing the methylation levels of populations of regions with different CNA characteristics. For example, the methylation level of regions corresponding to excessive abundance (e.g., Z _CNA > 3) can be compared with the methylation level of regions corresponding to normal abundance (e.g., -3 < Z _CNA < 3). First, the effect of CNA on methylation levels is described.

A. Changes in methylation density in regions with chromosome addition and loss.

Because tumor tissue generally exhibits overall hypomethylation, the presence of tumor-derived DNA in the plasma of cancer patients will cause a decrease in methylation density compared to non-cancer individuals. Theoretically, the degree of hypomethylation in the plasma of cancer patients is proportional to the percentage concentration of tumor-derived DNA in the plasma sample.

For regions of tumor tissue exhibiting chromosomal amplification, an additional dose of tumor DNA will be released from the amplified DNA segments into the plasma. This contribution of increased tumor DNA to the plasma will theoretically cause a higher degree of hypomethylation in the plasma DNA of the diseased region. Another factor is that the anticipated amplification of the genomic region will give tumor cells a growth advantage, thus presumably leading to its expression. Such regions are generally hypomethylated.

In contrast, for regions of tumor tissue exhibiting chromosomal loss, the reduced contribution of tumor DNA to plasma will result in a lower degree of hypomethylation compared to regions without copy number alterations. Another factor is that the missing genomic regions in tumor cells may contain tumor suppressor genes, potentially silencing these regions. Therefore, such regions are expected to have a higher probability of hypermethylation.

Here, results from two HCC patients (TBR34 and TBR36) are used to illustrate this effect. Figures 48A (TBR36) and 49A (TBR34) show regions of increased or decreased chromosomes highlighted by circles and the corresponding methylation analyses. Figures 48B and 49B show box plots of methylation z-scores for decreased, normal, and increased methylation in patients TBR36 and TBR34, respectively.

Figure 48A shows a Circos plot confirming changes in CNA (inner loop) and methylation (outer loop) in bisulfite-treated plasma DNA of HCC patient TBR36. Red circles highlight regions with chromosome increase or loss. Regions showing chromosome increase are more hypomethylated than regions without copy number changes. Regions showing chromosome loss are less hypomethylated than regions without copy number changes. Figure 48B is a curve showing the methylation z-score for regions with chromosome increase and loss, and regions without copy number changes, in HCC patient TBR36. Compared to regions without copy number changes, regions with chromosome increase have larger negative z-scores (more hypomethylated), and regions with chromosome loss have smaller negative z-scores (less hypomethylated).

Figure 49A shows a Circos plot confirming CNA (inner loop) and methylation alterations (outer loop) in bisulfite-treated plasma DNA from HCC patient TBR34. Figure 49B is a box plot of methylation z-fractions in regions with chromosomal increases and losses, as well as regions without copy number alterations, from HCC patient TBR34. The difference in methylation density between regions with chromosomal increases and losses was greater in patient TBR36 than in patient TBR34 because the former patient had a higher percentage concentration of tumor-derived DNA.

In this example, the region used to determine CNA is the same as the region used to determine methylation. In one embodiment, the threshold for the corresponding region depends on whether the corresponding region shows deletion or amplification. In one implementation, the magnitude of the threshold for the corresponding region (e.g., the z-score threshold for determining hypomethylation) is larger when the corresponding region shows amplification compared to when no amplification is shown (e.g., the magnitude can exceed 3, and a threshold below -3 can be used). Therefore, to test for hypomethylation, the threshold for the corresponding region can have a larger negative value when the corresponding region shows amplification compared to when no amplification is shown. It is expected that such implementations will improve the specificity of tests for detecting cancer.

In another embodiment, when a region shows a deletion, the threshold for that region has a smaller value (e.g., below 3) compared to when no deletion is shown. Therefore, to test for hypomethylation, the threshold for a region showing a deletion can have a smaller negative value compared to when no deletion is shown. Such implementations are expected to improve the sensitivity of tests used to detect cancer. The threshold adjustments in the above implementations can depend on the desired changes in sensitivity and specificity for a particular diagnostic situation. In other embodiments, methylation and CNA measurements can be used in conjunction with other clinical parameters (e.g., imaging results or serum biochemistry) to predict cancer.

B. Use CNA to select the region

As described above, it has been shown that plasma methylation density changes in regions with copy number abnormalities within tumor tissue. In regions of increased copy number within tumor tissue, the increased contribution of hypomethylated tumor DNA to plasma leads to a greater degree of plasma DNA hypomethylation compared to regions without copy number abnormalities. Conversely, in regions of copy number loss within tumor tissue, the reduced contribution of hypomethylated cancer-derived DNA to plasma leads to a lesser degree of plasma DNA hypomethylation. This relationship between plasma DNA methylation density and relative presentation can be used to distinguish hypomethylation results corresponding to the presence of cancer-related DNA from hypomethylation corresponding to other non-cancerous causes in plasma DNA (e.g., SLE).

To illustrate this method, plasma samples from two patients with hepatocellular carcinoma (HCC) and two patients with SLE but not cancer were analyzed. The two SLE patients (SLE04 and SLE10) showed significant hypomethylation and CNA in their plasma. For patient SLE04, 84% of the area showed hypomethylation and 11.2% showed CNA. For patient SLE10, 10.3% of the area showed hypomethylation and 5.7% showed CNA.

Figures 50A and 50B show the results of plasma hypomethylation and CNA analysis in SLE patients SLE04 and SLE10. The outer circle shows the methylation z-fraction (Z _-methylation ) at 1 Mb resolution. Regions with Z _-methylation < -3 are shown in red, and regions with Z _-methylation > -3 are shown in gray. The inner circle shows the CNA z-fraction (Z- _CNA ). Green, red, and gray dots represent regions with Z- _CNA >3, <3, and between -3 and 3, respectively. Hypomethylation and CNA alterations were observed in the plasma of both SLE patients.

To determine whether changes in methylation and CNA are consistent with the presence of cancer-derived DNA in plasma, Z- _methylation was compared in regions with Z- _CNA >3, <-3, and between -3 and 3. For changes in methylation and CNA caused by cancer-derived DNA in plasma, regions with Z _-CNA <-3 were expected to correspond to less hypomethylation and smaller negative Z _-methylation values. In contrast, regions with Z _-CNA >3 were expected to correspond to more hypomethylation and larger negative Z _-methylation values. For illustration, a one-sided rank-sum test was applied to compare Z-methylation in regions with CNA (i.e., regions with Z- _CNA <-3 or >3) with regions without CNA (i.e., _regions with Z- _CNA between -3 and 3). In other embodiments, other statistical tests may be used, such as (but not limited to) Stourden's t-test, analysis of variance (ANOVA) test, and Kruskal-Wallis test.

Figures 51A and 51B show Z- _methylation analysis of CNA-containing and CNA-free regions in plasma from two HCC patients (TBR34 and TBR36). Regions with Z _CNA <-3 and >3 represent regions of insufficient and excessive DNA presentation in plasma, respectively. In both TBR34 and TBR36, regions of insufficient presentation (i.e., Z _CNA <-3) in plasma showed significantly higher Z- _methylation compared to regions with normal presentation (i.e., regions with Z _CNA between -3 and 3) (P < ^10⁻⁵ , one-sided rank-sum test). Normal presentation corresponds to what is expected in euploid genomes. For regions of excessive presentation (i.e., Z _CNA >3), they showed significantly lower Z- _methylation compared to regions with normal presentation (P < ^10⁻⁵ , one-sided rank-sum test). All these changes are consistent with the presence of hypomethylated tumor DNA in plasma samples.

Figures 51C and 51D show Z- _methylation analysis of plasma regions with and without Z-CNA in two SLE patients (SLE04 and SLE10). Regions with Z _-CNA <-3 and >3 represent regions with insufficient and excessive Z-CNA in plasma, respectively. For SLE04, regions with insufficient Z _-CNA (i.e., Z-CNA <-3) did not have significantly higher Z _-methylation compared to regions with normal Z-CNA (i.e., regions with Z _-CNA between -3 and 3) (P = 0.99, one-sided rank-sum test), and regions with excessive Z-CNA (i.e., regions with Z _-CNA >3) did not have significantly lower Z- _methylation compared to regions with normal Z-CNA (P = 0.68, one-sided rank-sum test). These results differ from the expected changes due to the presence of tumor-derived hypomethylated DNA in plasma. Similarly, for SLE10, regions with Z _CNA < -3 did not have significantly higher Z _methylation compared to regions with Z _CNA between -3 and 3 (P = 0.99, one-sided rank-sum test).

The reason why the typical cancer-related pattern between Z _-methylation and Z _-CNA is not present in SLE patients is that CNA is absent in certain cell types that also exhibit hypomethylation in SLE patients. In fact, the observed presence of CNA and hypomethylation on the surface is due to altered size distribution of circulating DNA in SLE patients. When the reference is derived from healthy individuals, this altered size distribution may change the sequencing read density of different genomic regions, leading to the presence of CNA on the surface. As described in previous chapters, there is a correlation between the size of circulating DNA fragments and their methylation density. Therefore, altered size distribution can also lead to aberrant methylation.

Although regions with Z _CNA > 3 have slightly lower methylation levels than regions with Z _CNA between -3 and 3, the p-values for comparison are much higher than those observed in two cancer patients. In one embodiment, the p-value can be used as a parameter to determine the likelihood of a test case having cancer. In another embodiment, the difference in Z _methylation between regions with normal and abnormal presentation can be used as a parameter indicating the likelihood of cancer presence. In one embodiment, a group of cancer patients can be used to establish the correlation between Z _methylation and Z _CNA and to determine thresholds for different parameters to indicate changes consistent with the presence of cancer-derived hypomethylated DNA in the test plasma sample.

Therefore, in one embodiment, CNA analysis can be performed to determine a first group of regions that all show one of the following: deletion, amplification, or normal presentation. For example, the first group of regions may all show deletion, or all show amplification, or all show normal presentation (e.g., regions with a normal first amount, such as normal Z- _methylation ). The methylation level of this first group of regions can be determined (e.g., the first methylation level of method 2800 may correspond to the first group of regions).

CNA analysis can identify a second group of regions that all show one of the following: deletion, amplification, or normal presentation. The second group of regions will show a different appearance than the first group. For example, if the first group of regions is normal, then the second group of regions may show deletion or amplification. The level of second methylation can be calculated based on the corresponding number of methylated DNA molecules at the sites in the second group of regions.

A parameter can then be calculated between the first and second methylation levels. For example, a difference or ratio can be calculated and compared to a threshold. The difference or ratio can also be subjected to a probability distribution (e.g., as part of a statistical test) to determine the probability of obtaining the value, and this probability can be compared to a threshold to determine the cancer grade based on the methylation level. Such a threshold can be selected to distinguish samples with cancer from those without cancer (e.g., SLE).

In one embodiment, the methylation level of a first set of regions or mixtures of regions (i.e., a mixture showing amplified, deleted, and normal regions) can be determined. This methylation level can then be compared to a first threshold as part of a first-stage analysis. If the threshold is exceeded, this indicates the likelihood of cancer, and the above analysis can then be performed to determine whether the indication is a false positive. Thus, the final classification of the cancer grade may include a comparison of two methylation level parameters with a second threshold.

The first methylation level can be a statistical value (e.g., mean or median) of the regional methylation level calculated for each region in the first group of regions. The second methylation level can also be a statistical value of the regional methylation level calculated for each region in the second group of regions. As an example, the statistical value can be determined using a one-sided rank-sum test, Stewart's t-test, analysis of variance (ANOVA), or the Kruskal-Wallis test.

XI. Classification of Cancer Types

In addition to determining whether an organism has cancer, the examples can also identify the type of cancer associated with a sample. This identification of cancer type can use a pattern of overall hypomethylation, CpG island hypermethylation, and/or CNA. The pattern may include clustering patients with a known diagnosis using measured regional methylation levels, the corresponding CNA values for the regions, and the methylation levels of the CpG islands. The following results show that organisms with similar types of cancer have similar values for regions and CpG islands, as do non-cancer patients. During clustering, each value for a region or island can be a separate scale in the clustering process.

It is known that cancers of the same type share similar genetic and epigenetic alterations (E. Gebhart et al. 2004 Cytogenet Genome Res; 104:352-358; PA. Jones et al. 2007 Cell; 128:683-692). Below, we describe how patterns of CNA and methylation alterations detected in plasma can be used to infer the origin or type of cancer. Plasma DNA samples from HCC patients, non-HCC patients, and healthy controls were classified using, for example, hierarchical clustering analysis. Analysis was performed using, for example, the heatmap.2 function from the R scripting package (cran.r-project.org/web/packages/gplots/gplots.pdf).

To illustrate the potential of this method, two sets of standards (Group A and Group B) were used as examples to identify features suitable for classifying plasma samples (see Table 6). In other embodiments, other standards may be used to identify the features. The features used include overall CNA at 1 Mb resolution, overall methylation density at 1 Mb resolution, and CpG island methylation.

Table 6

In the first two examples, CNA, global methylation at 1 Mb resolution, and CpG island methylation features were all used for classification. In other embodiments, other criteria may be used, such as (but not limited to) measuring the accuracy of features in the plasma of a reference population.

Figure 52A illustrates a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using all 1,130 Group A features, including 355 CNAs, 584 global methylation features at 1 Mb resolution, and methylation status of 110 CpG islands. The top bands represent sample groups: green, blue, and red represent healthy individuals, HCC patients, and non-HCC cancer patients, respectively. Generally, the three groups tend to cluster together. The vertical axis represents categorical features. Features with similar patterns across different individuals cluster together. These results suggest that patterns of CpG island methylation alterations, genome-wide methylation alterations at 1 Mb resolution, and CNAs in plasma may be used to determine the origin of cancer in patients with unknown origins.

Figure 52B illustrates a hierarchical clustering analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using all 2,780 Group B features, including 759 CNAs, 911 whole-genome methylation states at 1 Mb resolution, and 191 CpG islands. The top bands represent sample groups: green, blue, and red represent healthy individuals, HCC patients, and non-HCC cancer patients, respectively. Generally, the three groups tend to cluster together. The vertical axis represents the categorical features. Features with similar patterns across different individuals cluster together. These results suggest that patterns of CpG island methylation alterations, whole-genome methylation alterations at 1 Mb resolution, and CNAs in different groups of plasma can be used to determine the origin of cancer in patients with unknown origins. The selection of categorical features can be tailored to specific applications. Furthermore, cancer type predictions can be weighted differently based on an individual's prior probability of different cancer types. For example, patients with chronic viral hepatitis tend to develop hepatocellular carcinoma, while chronic smokers tend to develop lung cancer. Therefore, weighted probabilities of cancer types can be calculated using, for example (but not limited to), logistic regression, multiple regression, or cluster regression.

In other embodiments, a single type of feature can be used for classification analysis. For example, in the following instances, only global methylation, CpG island hypermethylation, or CNA at 1Mb resolution is used for hierarchical clustering analysis. Discriminative power may vary when different features are used. Further optimization of classification features can improve classification accuracy.

Figure 53A illustrates a stratified clustering analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using group A CpG island methylation features. Generally, cancer patients clustered together, while non-cancer individuals were in another group. However, compared to using all three types of features, HCC and non-HCC patients were less clearly distinguishable.

Figure 53B shows a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls, using the overall methylation density of group A at 1 Mb resolution as a classification feature. Preferred clustering was observed between HCC and non-HCC patients.

Figure 54A shows a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls, using the overall CNA at 1 Mb resolution as the classification feature in group A. Preferred clustering is observed between HCC and non-HCC patients.

Figure 54B illustrates a hierarchical cluster analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls, using the methylation density of CpG islands in group B as a classification feature. Preferred clustering of HCC and non-HCC cancer patients can be observed.

Figure 55A illustrates a hierarchical clustering analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls, using the overall methylation density of group B at 1 Mb resolution as a classification feature. Preferred clustering of HCC and non-HCC cancer patients can be observed.

Figure 55B illustrates a hierarchical clustering analysis of plasma samples from HCC patients, non-HCC cancer patients, and healthy controls using the overall CNA of group B at 1 Mb resolution as a classification feature. Preferred clustering of HCC and non-HCC cancer patients can be observed.

These hierarchical clustering results from plasma samples indicate that combinations of different features can be used to identify major cancer types. Further optimization of the selection criteria could further improve classification accuracy.

Therefore, in one embodiment, when methylation classification indicates the presence of cancer in an organism, the type of cancer associated with the organism can be identified by comparing methylation levels (e.g., first methylation from method 2800 or methylation levels in any region) with corresponding values determined by other organisms (i.e., other organisms of the same type, such as humans). The corresponding values can be for the same set of regions or sites for which methylation levels are calculated. At least two of the other organisms are identified as having different types of cancer. For example, the corresponding values can be organized into classes, where two classes are associated with different cancers.

Furthermore, when CNA and methylation are used together to obtain a third category of cancer classification, CNA and methylation features can be compared with corresponding values from other organisms. For example, the first amount showing a region of deletion or amplification (e.g., from Figure 36) can be compared with corresponding values determined by other organisms to identify the cancer type associated with that organism.

In some embodiments, methylation characteristics are the regional methylation levels of multiple regions of the genome. Regions whose methylation levels exceed a corresponding regional threshold can be used; for example, the regional methylation levels of an organism can be compared with the regional methylation levels of the same regions in the genomes of other organisms. This comparison can allow for differentiation of cancer types, or simply provide additional filtering to confirm cancer (e.g., identifying false positives). Therefore, based on the comparison, it can be determined whether an organism has type I cancer, does not have cancer, or has type II cancer.

Other organisms (along with the tested organism) can be clustered using regional methylation levels. Therefore, comparisons of regional methylation levels can be used to determine which group an organism belongs to. Clustering can also use CNA normalized values to identify regions showing deletions or amplifications, as described above. Furthermore, clustering can use the corresponding methylation density of highly methylated CpG islands.

To illustrate the principle of this method, an example of classifying two unknown samples using logistic regression is presented. The purpose of this classification is to determine whether the two samples are HCC or non-HCC cancers. A training group of samples was compiled, which included 23 plasma samples collected from HCC patients and 18 samples from patients with cancers other than HCC. Therefore, there were a total of 41 cases in the training group. In this example, 13 features were selected, including five features (X1-X5) regarding CpG island methylation, six features (X6-X11) regarding methylation of the 1Mb region, and two features (X12-X13) regarding CNA in the 1Mb region. The CpG methylation feature was selected based on a z-score >3 or <-3 for at least 15 cases in the training group. The 1Mb methylation feature was selected based on a z-score >3 or <-3 for at least 39 cases in the training group. The CNA feature was selected based on a z-score >3 or <-3 for at least 20 cases. Logistic regression was performed on the samples from this training group to determine the regression coefficients for each feature (X1–X13). Features with larger regression coefficients (regardless of whether they were positive or negative) provided better differentiation between HCC and non-HCC samples. The z-score of the corresponding feature for each case was used as input values for the independent variables. Thirteen features from two plasma samples were then analyzed, one from an HCC patient (TBR36) and the other from a patient with lung cancer (TBR177).

In this cancer type classification analysis, it is assumed that both samples were collected from patients with cancer of unknown origin. For each sample, the z-score of the corresponding feature is put into a logistic regression equation to determine the natural logarithm of the odds ratio (ln(concession ratio)), where the concession ratio represents the ratio of the probability of having HCC to the probability of not having HCC (HCC/non-HCC).

Table 7 shows the regression coefficients for the 13 features of the logistic regression equation. The z-scores for the corresponding features for the two test cases (TBR36 and TBR177) are also shown. The ln(odds ratio) for HCC in TBR36 and TBR177 are 37.03 and -4.37, respectively. From these odds ratios, the probabilities of collecting plasma samples from HCC patients are >99.9% and 1%, respectively. Simply put, TBR36 has a high probability of the sample originating from an HCC patient, while TBR177 has a low probability.

Table 7

In other embodiments, hierarchical clustering regression, classification tree analysis, and other regression models can be used to determine the likely primary source of cancer.

XII. Materials and Methods

A. Preparation of bisulfite-treated DNA libraries and sequencing

Genomic DNA with 0.5% (w/w) unmethylated λDNA (Promega) was fragmented to approximately 200 bp using the Covaris S220 system. DNA libraries were prepared according to the manufacturer's instructions using a paired-end sequencing sample preparation kit (Ilumina), except that methylated adaptors (Ilumina) were attached to the DNA fragments. After two rounds of purification using AMPure XP magnetic beads (Beckman Coulter), the conjugation product was aliquoted into two fractions, one of which underwent two rounds of bisulfite modification using the EpiTect bisulfite kit (Qiagen). Unmethylated cytosine at CpG sites in the insert was converted to uracil, while methylated cytosine remained unchanged. DNA molecules conjugated by adaptors, whether treated with sodium bisulfite or not, were enriched by 10 PCR cycles using the following formulation: 50 μl of 2.5 U PfuTurboCx hotstart DNA polymerase (Agilent Technologies), 1X PfuTurboCx reaction buffer, 25 μM dNTP, 1 μl of PCR primer PE 1.0, and 1 μl of PCR primer PE 2.0 (Illumina). The thermal cycling pattern was: 95 °C for 2 min, 98 °C for 30 s, followed by 10 cycles of 98 °C for 15 s, 60 °C for 30 s, and 72 °C for 4 min, and a final step of 72 °C for 10 min (R. Lister, et al. 2009 Nature; 462:315-322). PCR products were purified using Amprey XP magnetic beads.

Plasma DNA was extracted from 3.2–4 ml maternal plasma samples, and fragmented λDNA (25 pg per ml of plasma) was added, followed by library construction as described above (Qiu et al., 2011 BMJ; 342:c7401). After conjugation to a methylated linker, the conjugation product was split in two, and one half underwent two rounds of bisulfite modification. Subsequently, the bisulfite-treated or untreated conjugation products were enriched by 10 PCR cycles as described above.

DNA libraries, either bisulfite-treated or untreated, were sequenced in paired-end format targeting 75 bp on a HiSeq2000 instrument (Illumina). DNA clusters were generated on a cBot instrument (Illumina) using a paired-end cluster generation kit v3. Real-time image analysis and base determination were performed using HiSeq Control Software (HCS) version 1.4 and Real-Time Analysis (RTA) software v1.13 (Illumina). These software programs automated matrix and phasing calculations based on the external DNA library sequencing using PhiX Control v3.

B. Sequence alignment and identification of methylated cytosine

After base determination, linker sequences and low-quality bases (i.e., mass fraction <20) at the fragment ends are removed. The FASTQ formatted trimmed reads are then processed through a methylation data analysis pipeline called Methy-Pipe (Jiang et al., Methy-Pipe: An integrated bioinformatics data analysis pipeline for whole genome methylome analysis, paper presented at the IEEE International Conference on Bioinformatics and Biomedicine Workshops, Hong Kong, 18-21 December 2010). To align bisulfite-converted sequencing reads, a computer program was first used to convert all cytosine residues to thymine separately in the Watson and Crick chains using a reference human genome (NCBI build 36/hg18). Subsequently, a computer program was used to convert each cytosine to thymine in all treated reads, and the positional information of each converted residue was saved. The converted reads were aligned to two converted reference human genomes using SOAP2 (Li et al. 2009 Bioinformatics 25:1966-1967), with a maximum of two mismatches allowed per aligned read. Only reads that could be aligned to a unique position in the genome were selected. Indistinct reads that aligned to both the Watson and Crick chains and duplicate (clonal) reads with the same start and end genomic positions were removed. Sequencing reads with an insert size ≤600 bp were retained for methylation and size analysis.

Cytosine residues in the CpG dinucleotide background are the primary target for downstream DNA methylation studies. After alignment, cytosine residues initially present on the sequencing reads are recovered based on positional information preserved during computer-programmed conversion. Recovered cytosine residues in CpG dinucleotides are graded as methylated. Thymine residues in CpG dinucleotides are graded as unmethylated. Unmethylated λDNA included during library preparation serves as an internal control for evaluating sodium bisulfite modification. If the bisulfite conversion efficiency is 100%, then all cytosine residues on the λDNA should have been converted to thymine.

XIII. Overview

Using the embodiments described herein, cancer can be noninvasively screened, detected, monitored, or predicted using, for example, individual plasma. Prenatal screening, diagnosis, research, or monitoring of the fetus can also be performed by inferring the methylation pattern of fetal DNA from maternal plasma. To illustrate the capabilities of the methods, it is demonstrated that information routinely obtained from studying placental tissue can be directly assessed from maternal plasma. For example, the imprinting status of loci, identification of loci with methylation differences between fetal and maternal DNA, and gestational changes in the methylation pattern of loci were achieved through direct analysis of maternal plasma DNA. A key advantage of our method is that the fetal methylome can be comprehensively assessed during pregnancy without disrupting the pregnancy or requiring invasive sampling of fetal tissue. Assuming known associations between altered DNA methylation status and many pregnancy-related conditions, the methods described in this study can serve as an important tool for studying the pathophysiology of those conditions and identifying biomarkers. By focusing on imprinted loci, it is demonstrated that paternally and maternally transmitted fetal methylation patterns can be assessed from maternal plasma. This method is applicable to the study of imprinted diseases. The examples can also be used directly for prenatal assessment of fetal or pregnancy-related diseases.

Whole-genome bisulfite sequencing has been proven effective in studying DNA methylation patterns in placental tissues. The human genome contains approximately 28 million CpG sites (Clark et al., 2012 PLoS One; 7:e50233). Bisulfite sequencing data from CVS and full-term placental tissue samples cover more than 80% of the CpGs. This indicates a significantly wider coverage than that achievable using other high-throughput platforms. For example, the illumina-infinim human methylated 27K bead array used in previous studies on placental tissue (T Chu et al., 2011 PLoS One; 6:e14723) only covers 0.1% of the CpGs in the genome. The recently available ilumininfinium human methylated 450K bead array only covers 1.7% of CpG (Clark et al. 2012 PLOS ONE; 7:e50233). Because the MPS method has no limitations related to probe design, hybridization efficiency, or antibody capture strength, it is possible to evaluate CpG inside or outside CpG islands and in most sequence backgrounds.

XIV. Computer Systems

Any computer system mentioned herein can utilize any suitable number of subsystems. Examples of such subsystems are shown in computer device 3300 in FIG33. In some embodiments, the computer system includes a single computer device, wherein the subsystems may be components of the computer device. In other embodiments, the computer system may include multiple computer devices, each having internal components, each serving as a subsystem.

The subsystems shown in Figure 33 are interconnected via system bus 3375. Other subsystems are shown, such as printer 3374, keyboard 3378, storage device 3379, monitor 3376 coupled to display adapter 3382, etc. Peripheral devices and input/output (I/O) devices coupled to I/O controller 3371 can be connected to the computer system via many components known in the art, such as serial port 3377. For example, serial port 3377 or external interface 3381 (e.g., Ethernet, Wi-Fi, etc.) can be used to connect computer system 3300 to a wide area network (e.g., the Internet), a mouse input device, or a scanner. The interconnection via system bus 3375 enables central processing unit 3373 to communicate with each subsystem and control the execution of instructions from system memory 3372 or storage device 3379 (e.g., fixed disk) and the exchange of information between subsystems. System memory 3372 and/or storage device 3379 may contain computer-readable media. Any value mentioned in this article can be output from one component to another and can also be output to the user.

A computer system may include multiple identical components or subsystems connected together, for example, via an external interface 3381 or an internal interface. In some embodiments, the computer system, subsystem, or device may communicate via a network. In this case, one computer may be regarded as a client and another computer as a server, wherein each may be part of the same computer system. The client and server may each include multiple systems, subsystems, or components.

It should be understood that any embodiment of the present invention can be implemented in a modular or integrated manner using hardware (e.g., application-specific integrated circuits or field-programmable gate arrays) and/or computer software with a general-purpose programmable processor in the form of control logic. As used herein, the processor includes a multi-core processor on the same integrated chip, or multiple processing units on a single circuit board or network-connected. Based on the present invention and the teachings provided herein, those skilled in the art will know and understand other ways and/or methods of implementing embodiments of the present invention using hardware and combinations of hardware and software.

Any software component or function described in this application can be implemented as software code executed by a processor using any suitable computer language (e.g., Java, C++, or Perl) and employing techniques such as conventional or object-oriented methods. The software code can be stored as a series of instructions or commands on a computer-readable medium for storage and/or transmission, suitable media including random access memory (RAM), read-only memory (ROM), magnetic media (e.g., hard disk drive or floppy disk) or optical media (e.g., optical disc (CD) or DVD (Digital Universal Optical Disc)), flash memory, etc. The computer-readable medium can be any combination of such storage or transmission means.

The program can also be encoded and transmitted using carrier signals suitable for transmission via wired, optical, and/or wireless networks (including the Internet) conforming to various schemes. Therefore, a computer-readable medium according to an embodiment of the invention can be generated using data signals encoded with such a program. Computer-readable media encoded with program code can be packaged with a compatible device or provided separately from other devices (e.g., via Internet download). Any such computer-readable medium can reside on or within a single computer program product (e.g., hard disk drive, CD, or entire computer system) and can reside on or within different computer program products within a system or network. The computer system may include a monitor, printer, or other suitable display for providing a user with any of the results mentioned herein.

Any method described herein can be performed, wholly or partially, by a computer system including one or more processors configured to perform the steps. Therefore, embodiments may relate to computer systems configured to perform the steps of any method described herein, possibly with different components performing the respective steps or groups of steps. Although the steps of the methods herein are presented as numbered steps, they may be performed simultaneously or in different orders. Furthermore, portions of these steps may be used in conjunction with portions of other steps of other methods. Moreover, all or part of the steps may be optional. Additionally, any step of any method may be performed using modules, circuitry, or other components that perform these steps.

Specific details of a particular embodiment may be combined in any suitable manner without departing from the spirit and scope of the embodiments of the invention. However, other embodiments of the invention may relate to specific embodiments associated with each individual aspect or a specific combination of these individual aspects.

The foregoing description of exemplary embodiments of the invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms described, and many modifications and variations are possible given the foregoing teachings. The embodiments have been chosen and described to best explain the principles of the invention and its practical application, thereby enabling others skilled in the art to best utilize the invention in various embodiments and with various modifications suitable for the particular intended use.

Unless specifically indicated to the contrary, the phrase “a/an” or “the” is intended to mean “one or more”.

All patents, patent applications, publications, and descriptions mentioned herein are incorporated herein by reference in their entirety for all purposes. None of them are acknowledged as prior art.

Table S2A. List of 100 most methylated regions identified from early pregnancy chorionic villus samples and maternal blood cells.

 

 

Table S2B. List of 100 hypomethylated regions identified from early pregnancy chorionic villus samples and maternal blood cells.

Table S2C. List of 100 most methylated regions identified from placental samples and maternal blood cells in late pregnancy.

Table S2D. List of 100 minimally methylated regions identified from placental samples and maternal blood cells in late pregnancy.

Table S3A lists the top 100 loci inferred as hypermethylated from early pregnancy maternal plasma bisulfite sequencing data.

 

 

Table S3B lists the top 100 hypomethylated loci inferred from early pregnancy maternal plasma bisulfite sequencing data.

 

 

Table 1 lists the top 100 S3C loci inferred from late-pregnancy maternal plasma bisulfite sequencing data as hypermethylated.

 

 

Table S3D lists the top 100 loci inferred as hypomethylated from maternal plasma bisulfite sequencing data in late pregnancy.

 

 

Claims (234)

1. A method for analyzing a biological sample of an organism for non-diagnostic purposes, said biological sample comprising a mixture of free deoxyribonucleic acid (DNA) molecules derived from normal cells and cancer-associated cells, said method comprising: The analysis of the biological sample comprises the analysis of multiple cell-free DNA molecules, wherein the analysis is performed on sequence reads obtained from massively parallel sequencing, and the analysis of each of the multiple cell-free DNA molecules includes: Determine the location of free DNA molecules within the genome of the organism; and Determine whether the free DNA molecule is methylated at one or more sites; For each of the multiple sites: Using sequence reads, the number of free DNA molecules from the biological sample methylated at the stated site is counted separately. The first methylation level is calculated by summing the corresponding numbers of free DNA molecules methylated at the plurality of sites, wherein the plurality of sites are located on a plurality of chromosomes; Compare the first methylation level with a first threshold; and Based on the comparison, a first category of cancer severity is determined. 2. The method of claim 1, wherein determining whether the free DNA molecule is methylated at one or more sites comprises: Perform sequencing that identifies methylation. 3. The method of claim 2, wherein performing identifiable methylation sequencing comprises: The free DNA molecules were treated with sodium bisulfite; and The processed free DNA molecules were sequenced. 4. The method of claim 3, wherein the treatment of the free DNA molecule with sodium bisulfite is part of Tet-assisted bisulfite conversion or oxidative bisulfite sequencing for the detection of 5-hydroxymethylcytosine. 5. The method of claim 2, wherein the corresponding number of free DNA molecules is determined by comparing sequence reads obtained from sequencing of the recognizable methylation. 6. The method of claim 1, wherein determining whether the free DNA molecule is methylated at one or more sites comprises digestion with a methylation-sensitive restriction enzyme, methylation-specific PCR, methylation-dependent DNA precipitation, methylated DNA-binding proteins/peptides, or single-molecule sequencing without sodium bisulfite treatment. 7. The method of claim 1, further comprising: For each of the first plurality of regions of the genome: Determine the corresponding number of free DNA molecules from the region of the biological sample; Calculate the corresponding normalized value for the corresponding number of free DNA molecules from the region; and The corresponding standardized value is compared with the reference value to determine whether the corresponding region shows a missing or an amplified region. A first quantity is determined for a first plurality of regions of the genome that are identified as containing deletions or amplifications; The first quantity is compared with a first threshold to determine a second category of cancer severity; and A third category is used to determine the cancer grade using the first and second classifications. 8. The method of claim 7, wherein the first threshold is a percentage of the first plurality of regions that are missing or amplified. 9. The method of claim 7, wherein the third category is cancer-positive only if both the first category and the second category indicate cancer. 10. The method of claim 7, wherein the third category is cancer-positive when the first or second category indicates cancer. 11. The method of claim 1, wherein the first classification indicates the presence of cancer in the organism, the method further comprising: The type of cancer associated with the organism is identified by comparing the first methylation level with a corresponding value determined from other organisms, wherein at least two of the other organisms are identified as having different types of cancer. 12. The method of claim 7, wherein the first classification indicates the presence of cancer in the organism, the method further comprising: The type of cancer associated with the organism is identified by comparing the first methylation level with a corresponding value determined from other organisms, wherein at least two of the other organisms are identified as having different types of cancer. 13. The method of claim 12, wherein the third classification indicates the presence of cancer in the organism, the method further comprising: The type of cancer associated with the organism is identified by comparing the first quantity in the region with a corresponding value determined from the other organisms. 14. The method of claim 7, wherein calculating the first methylation level comprises: Identify a second or more regions of the genome; Identify one or more sites within each of the second plurality of regions; and Calculate the regional methylation level of each of the second plurality of regions, wherein the first methylation level is for the first region. The method further includes: Each of the methylation levels in a region is compared with a corresponding regional threshold, including comparing the first methylation level with the first threshold. A second amount is determined for regions where the methylation level exceeds the corresponding regional threshold; and The second quantity of a region in the second plurality of regions is compared with a second threshold to determine the first classification. 15. The method of claim 1, wherein calculating the first methylation level comprises: Identify a second or more regions of the genome; Identify one or more sites within each of the second plurality of regions; and Calculate the regional methylation level of each of the second plurality of regions, wherein the first methylation level is for the first region. The method further includes: Each of the methylation levels in a region is compared with a corresponding regional threshold, including comparing the first methylation level with the first threshold. A second amount is determined for regions where the methylation level exceeds the corresponding regional threshold; and The second quantity of a region in the second plurality of regions is compared with a second threshold to determine the first classification. 16. The method of claim 15, wherein the region where the methylation level exceeds the corresponding region threshold corresponds to a first group of regions, the method further comprising: The methylation levels of the first set of regions are compared with the corresponding methylation levels of other organisms in the first set of regions, wherein the other organisms have at least two of the following: type I cancer, no cancer, and type II cancer; and Based on the comparison, it is determined whether the organism has the first type of cancer, does not have cancer, or has the second type of cancer. 17. The method of claim 16, further comprising: The other organisms are clustered based on the methylation levels of corresponding regions in the first group of regions, wherein two of the clusters correspond to either of the following: the first type of cancer, the absence of cancer, and the second type of cancer. The comparison of regional methylation levels in the second plurality of regions is used to determine which category the organism belongs to. 18. The method of claim 17, wherein the clustering of the other organisms uses the methylation level of the regions of the organism. 19. The method of claim 17, wherein the class includes a first class corresponding to the first type of cancer, a second class corresponding to the second type of cancer, and a third class corresponding to the absence of cancer. 20. The method of claim 17, wherein the clustering of the other organisms is further based on corresponding normalized values of a second set of regions of the other organisms, wherein the second set of regions corresponds to regions identified as containing deletions or amplifications, and wherein the corresponding normalized value of the region is determined by the corresponding number of free DNA molecules from the region, the method further comprising: For each of the regions in the second group: Determine the corresponding number of free DNA molecules from the region; and The corresponding normalized value is calculated from the corresponding number of free DNA molecules from the region; and The corresponding standardized value of the second group region of the organism is compared with the corresponding standardized value of the other organisms to determine which category the organism belongs to. 21. The method of claim 20, wherein the clustering of the other organisms is further based on the corresponding methylation density of highly methylated CpG islands, the method further comprising: For each of the hypermethylated CpG islands: Determine the corresponding methylation density, and The methylation density of the hypermethylated CpG islands of the organism is compared with the methylation density of the other organisms to determine which category the organism belongs to. 22. The method of claim 15, further comprising: For each of the second plurality of regions: Calculate the difference between the methylation level of the region and the corresponding threshold of the region; and Calculate the corresponding probability for each of the aforementioned differences; The second quantity that determines the region includes: Calculate the cumulative score including the corresponding probability. 23. The method of claim 22, wherein calculating the cumulative score comprises: The logarithmic result is obtained by using the logarithm of the corresponding probability of each region in the second plurality of regions; and The calculation includes the sum of the corresponding logarithmic results. 24. The method of claim 23, wherein the cumulative fraction is the negative of the sum of the corresponding logarithmic results. 25. The method of claim 22, wherein the corresponding differences in each of the second plurality of regions are standardized by the standard deviation related to the threshold of the corresponding region. 26. The method of claim 22, wherein the corresponding probability corresponds to the probability of the corresponding difference according to a statistical distribution. 27. The method of claim 22, wherein the second threshold corresponds to the highest cumulative score of a reference group of samples from other organisms. 28. The method of claim 15, further comprising: For each of the first plurality of regions: Calculate the corresponding difference between the standardized value and the reference value; and Calculate the corresponding probability for each of the aforementioned differences; The first quantity that determines the region includes: The calculation includes the first summation of the corresponding probabilities. 29. The method of claim 15, wherein the corresponding region threshold is a specified amount from a reference methylation level. 30. The method of claim 15, wherein the second threshold is a percentage, and wherein comparing the second quantity in the comparison region with the second threshold comprises: Before comparing with the second threshold, the second quantity of the region is divided by the second number of the second plurality of regions. 31. The method of claim 30, wherein the second number corresponds to all of the second plurality of regions. 32. The method of claim 15, wherein the first plurality of regions are the same as the second plurality of regions, and wherein the threshold of the corresponding region depends on whether the corresponding region shows a deficiency or an amplification. 33. The method of claim 32, wherein one of the threshold values of the corresponding region when amplification is displayed has a larger value compared to when no amplification is displayed, and wherein the second of the threshold values of the corresponding region when a deficiency is displayed has a smaller value compared to when no deficiency is displayed. 34. The method of claim 33, wherein the corresponding region threshold tests for hypomethylation of the second plurality of regions, wherein the third of the corresponding region thresholds has a larger negative value when the corresponding region shows amplification compared to when no amplification is shown, and wherein the fourth of the corresponding region thresholds has a smaller negative value when the corresponding region shows deletion compared to when no deletion is shown. 35. The method of claim 15, wherein the biological sample is collected prior to treatment, the method further comprising: For another biological sample collected after treatment, the method described in claim 15 is repeated to obtain: Determine the first subsequent measure to show the region of absence or amplification; and A subsequent second quantity is determined for regions where the methylation level exceeds the corresponding regional threshold; The first quantity is compared with the subsequent first quantity, and the second quantity is compared with the subsequent second quantity to determine the prognosis of the organism. 36. The method of claim 35, wherein comparing the first quantity with the subsequent first quantity and comparing the second quantity with the subsequent second quantity to determine the prognosis of the organism comprises: Determine the first difference between the first quantity and the subsequent first quantity; Compare the first difference with one or more first difference thresholds; Determine the second difference between the second quantity and the subsequent second quantity; and The second difference is compared with one or more second difference thresholds. 37. The method of claim 36, wherein when the first difference is below one of the first difference thresholds, the prognosis is predicted to be worse compared to when the first difference exceeds one of the first difference thresholds, and wherein when the second difference is below one of the second difference thresholds, the prognosis is predicted to be worse compared to when the second difference exceeds one of the second difference thresholds. 38. The method of claim 37, wherein one of the first difference thresholds and one of the second difference thresholds are zero. 39. The method of claim 35, wherein the treatment is immunotherapy, surgery, radiation therapy, chemotherapy, antibody-based therapy, epigenetic therapy, or targeted therapy. 40. The method of claim 1, wherein the first threshold is a specified distance from a reference methylation level established from a biological sample obtained from a healthy organism. 41. The method of claim 40, wherein the specified distance is a specified number of standard deviations relative to the reference methylation level. 42. The method of claim 1, wherein the first threshold is established by a reference methylation level, the reference methylation level being determined from a previous biological sample of the organism obtained prior to the biological sample test. 43. The method of claim 1, wherein comparing the first methylation level with the first threshold comprises: Determine the difference between the first methylation level and the reference methylation level; and The difference is compared with the value corresponding to the first threshold. 44. The method of claim 1, further comprising: Determine the percentage concentration of cell-free tumor DNA in the biological sample; The first threshold is calculated based on the percentage concentration. 45. The method of claim 1, further comprising: Determine whether the percentage concentration of cell-free tumor DNA in the biological sample exceeds a minimum value; and If the percentage concentration does not exceed the minimum value, then the biological sample is labeled. 46. The method of claim 45, wherein the minimum value is determined based on the expected difference between the tumor methylation level and a reference methylation level. 47. The method of claim 1, further comprising: Measuring the size of free DNA molecules located at the plurality of sites; and Before comparing the first methylation level with the first threshold, the first methylation level is normalized based on the measured size of the free DNA molecule. 48. The method of claim 47, wherein normalizing the first methylation level based on the measurement size comprises: Select free DNA molecules with the first size; The first methylation level is calculated using the selected free DNA molecule, and the first threshold corresponds to the first size. 49. The method of claim 48, wherein the first dimension is a length range. 50. The method of claim 48, wherein the free DNA molecules are selected based on size-dependent physical separation. 51. The method of claim 48, wherein selecting a free DNA molecule having a first size comprises: The plurality of free DNA molecules were subjected to paired-end massive parallel sequencing to obtain the paired sequences of each of the plurality of free DNA molecules; The size of the free DNA molecules is determined by comparing the paired sequences of the plurality of free DNA molecules with a reference genome; and Select a free DNA molecule having the first size. 52. The method of claim 47, wherein normalizing the first methylation level based on the measured size comprises: Obtain the functional relationship between size and methylation level; and The first methylation level is normalized using the aforementioned functional relationship. 53. The method of claim 52, wherein the functional relationship provides a correction coefficient corresponding to the respective size. 54. The method of claim 53, further comprising: Calculate the average size corresponding to the free DNA molecule used to calculate the first methylation level; and Multiply the first methylation level by the corresponding correction factor. 55. The method of claim 53, further comprising: For each of the plurality of sites: For each of the free DNA molecules located at the said site: Obtain the corresponding size of the free DNA molecule at the said site; and The contribution of the free DNA molecule to the corresponding number of free DNA molecules methylated at the site is normalized using the correction coefficient corresponding to the corresponding size. 56. The method of claim 1, wherein the plurality of sites comprises CpG sites, wherein the CpG sites are organized into a plurality of CpG islands, each CpG island comprising a plurality of CpG sites, and wherein the first methylation level corresponds to a first CpG island of the plurality of CpG islands. 57. The method of claim 56, wherein each of the CpG islands in a reference population of samples from other organisms has an average methylation density of less than a first percentage, and wherein each of the CpG islands in the reference population has an average methylation density coefficient of variation of less than a second percentage, and wherein for each CpG island, each methylation density is determined using the total number of methylated and unmethylated DNA molecules spanning the plurality of CpG sites. 58. The method of claim 56, further comprising: For each of the CpG islands: By comparing the methylation level of the CpG islands with corresponding thresholds, it is determined whether the CpG islands are highly methylated relative to a reference group of samples from other organisms; For each of the hypermethylated CpG islands: Determine the corresponding methylation density, Calculate the cumulative fraction from the corresponding methylation density; and The cumulative score is compared with a cumulative threshold to determine the first classification. 59. The method of claim 58, wherein calculating the cumulative fraction from the corresponding methylation density comprises: For each of the hypermethylated CpG islands: Calculate the corresponding difference between the methylation density and the reference density; and Calculate the corresponding probability corresponding to the corresponding difference; and The cumulative score is determined using the corresponding probability. 60. The method of claim 59, wherein the cumulative score is determined by: For each of the hypermethylated CpG islands: the corresponding logarithmic result is obtained using the logarithm of the corresponding probability; and The calculation includes the sum of the corresponding logarithmic results, wherein the cumulative fraction is the negative of the sum. 61. The method of claim 59, wherein each corresponding difference is standardized using a standard deviation associated with the reference density. 62. The method of claim 58, wherein the cumulative threshold corresponds to the highest cumulative score from the reference group. 63. The method of claim 58, wherein determining whether the first CpG island is hypermethylated comprises: Compare the first methylation level with the first threshold and the third threshold. The first threshold corresponds to the average methylation density of the reference population plus a specified percentage, and the third threshold corresponds to a specified number of standard deviations plus the average methylation density of the reference population groups. 64. The method of claim 63, wherein the specified percentage is 2%. 65. The method of claim 63, wherein the specified number of standard deviations is three. 66. The method of claim 1, further comprising: For each of the first plurality of regions of the genome: Determine the corresponding number of free DNA molecules from the region; Calculate the corresponding standardized value from the corresponding number; and The corresponding standardized value is compared with the reference value to determine whether the region shows a missing or expanded feature. A first group of regions is identified, which is defined as having one of the following: deletion, amplification, or normal presentation, wherein the first methylation level corresponds to the first group of regions; A second group of regions was identified, which was defined as having another of the following: deletion, amplification, or normal presentation; and The second methylation level is calculated based on the corresponding number of free DNA molecules methylated at the sites in the second group of regions. The comparison of the first methylation level with the first threshold includes: Calculate the parameters between the first methylation level and the second methylation level; and The parameter is compared with the first threshold. 67. The method of claim 66, wherein the first methylation level is a statistical value of the regional methylation level calculated for each region of the first group of regions, and wherein the second methylation level is a statistical value of the regional methylation level calculated for each region of the second group of regions. 68. The method of claim 67, wherein the statistic is determined using a Student’s t-test, an analysis of variance (ANOVA) test, or a Kruskal-Wallis test. 69. The method of claim 66, wherein the parameter includes the ratio or difference between the first methylation level and the second methylation level. 70. The method of claim 69, wherein calculating the parameter includes applying a probability distribution to the ratio or the difference. 71. A method for non-diagnostic purposes of determining a first methylation pattern from a biological sample of an organism, said biological sample comprising cell-free DNA containing a mixture of cell-free DNA molecules derived from a first tissue and a second tissue, said method comprising: A second methylation pattern corresponding to the DNA molecules of the second tissue is obtained, the second methylation pattern providing a second methylation density at each of a plurality of loci in the genome of the organism, the second methylation density at each locus being the proportion of methylated DNA molecules of the second tissue at the locus; The free methylation pattern is determined from the free DNA molecules of the mixture, the free methylation pattern provides the mixed methylation density at each of the plurality of loci, the mixed methylation density at each locus being the proportion of methylated DNA molecules of the mixture at that locus; Determine the percentage of the free DNA molecules from the first tissue in the mixture; and The first methylation type of the first tissue is determined by the following: For each of the plurality of loci: The difference parameter is calculated, which includes the difference between the second methylation density of the second methylation type and the mixed methylation density of the free methylation type, the difference being measured by the percentage of the free DNA molecules from the first tissue in the mixture. 72. The method of claim 71, further comprising: Determine the first methylation level from the first methylation morphology; and The first methylation level is compared with a first threshold to determine the cancer grade classification. 73. The method of claim 71, further comprising: The first methylation type is transformed to obtain the corrected first methylation type. 74. The method of claim 73, wherein the transformation is a linear transformation. 75. The method of claim 71, wherein the differential parameter D of the locus is defined as follows: mbc represents the second methylation density of the second methylation pattern at the locus, mp represents the mixed methylation density of the free methylation pattern at the locus, f is the percentage concentration of free DNA molecules from the first tissue in the biological sample, and CN represents the copy number at the locus. 76. The method of claim 75, wherein CN is one for the copy number of the first tissue that is diploid at the locus. 77. The method of claim 75, further comprising: Identify regions where D exceeds the threshold. 78. The method of claim 71, further comprising: The plurality of loci are identified by selecting loci that meet one or more of the following criteria: GC content exceeding 50%; The second methylation density of the second methylation type is below the first threshold or exceeds the second threshold; and The region defined by the locus contains at least five CpG sites. 79. The method of claim 71, wherein the biological sample is derived from a female individual carrying a fetus, and wherein the first tissue is derived from the fetus or placenta and the second tissue is derived from the female individual. 80. The method of claim 79, further comprising: Determine the first methylation level from the first methylation morphology; and The first methylation level is compared with a first threshold to determine the classification of the medical condition. 81. The method of claim 80, wherein the medical condition is preeclampsia or a chromosomal abnormality in the first tissue. 82. The method of claim 71, wherein the first tissue is derived from a tumor within the organism and the second tissue is derived from a non-malignant tissue within the organism. 83. The method of claim 71, wherein the biological sample further comprises DNA molecules from the second tissue, and wherein obtaining the second methylation type comprises: The DNA molecules from the second tissue are analyzed to determine the second methylation pattern. 84. The method of claim 71, wherein the obtained second methylation type corresponds to the average methylation type obtained from a set of control samples. 85. The method of claim 71, wherein determining the mixed methylation density of the methylation patterns of the free DNA at the first locus of the plurality of loci comprises: Determine a first number of the free DNA molecules from the first locus; Determine whether each of the first number of DNA molecules is methylated at one or more sites to obtain a second number of methylated DNA molecules; and The mixed methylation density is calculated from the first number and the second number. 86. The method of claim 85, wherein the one or more sites are CpG sites. 87. The method of claim 71, wherein determining the free DNA methylation pattern comprises: The free DNA molecules were treated with sodium bisulfite; and The processed DNA molecules were sequenced. 88. The method of claim 87, wherein the treatment of the free DNA molecule is part of a Tet-assisted bisulfite conversion or includes treatment of the free DNA molecule with potassium perruthenate (KRuO4). 89. The method of claim 71, wherein determining the free methylation type comprises: The DNA is subjected to single-molecule sequencing, wherein the single-molecule sequencing includes determining whether at least one site of the DNA molecule is methylated. 90. A computer system comprising one or more processors and a computer-readable medium storing a plurality of instructions, which, when executed, control one or more processors of the computer system to perform the method as described in any one of claims 1 to 6-89. 91. A method for non-diagnostic purposes of determining a first methylation pattern from a biological sample of an organism, said biological sample comprising free DNA containing a mixture of free DNA molecules derived from a first tissue and a second tissue, said method comprising: Analyzing multiple DNA molecules from the biological sample, wherein the DNA molecules analyzed from the multiple DNA molecules include: Determine the location of the DNA molecule in the genome of the organism; Determine the genotype of the DNA molecule; and Determine whether the DNA molecule is methylated at one or more sites; Identify multiple first loci that satisfy the condition that the first allele and the corresponding second allele corresponding to the first genome of the first tissue are heterozygous and the corresponding first allele of the second genome of the second tissue is homozygous; For each of the first loci: For each of the one or more loci associated with the said gene locus: Determine the number of DNA molecules methylated at the site and corresponding to the corresponding second allele at the locus; The methylation density is calculated using the number of DNA molecules methylated at one or more sites at the locus and corresponding to the respective second allele at the locus; and The first methylation pattern of the first tissue is generated from the methylation density of the first locus, wherein the first methylation density at each locus is the proportion of methylated DNA molecules of the first tissue at the locus. 92. The method of claim 91, wherein the biological sample is derived from a female individual carrying a fetus, and wherein the first tissue is derived from the fetus or placenta and the second tissue corresponds to the female individual. 93. The method of claim 91, wherein the site corresponds to a CpG site. 94. The method of claim 91, wherein each locus includes at least one CpG site. 95. The method of claim 91, wherein each locus is adjacent to at least one CpG site. 96. A method for detecting chromosomal abnormalities from a biological sample of an organism for non-diagnostic purposes, said biological sample comprising cell-free DNA containing a mixture of cell-free DNA molecules derived from a first tissue and a second tissue, said method comprising: The analysis comprises multiple DNA molecules from the biological sample, wherein the analysis is performed on sequence reads obtained from massively parallel sequencing, and the DNA molecules analyzed from the multiple DNA molecules include: Determine the location of the DNA molecule described in the reference genome; and Determine whether the DNA molecule is methylated at one or more sites; For each of the multiple sites: The number of DNA molecules methylated at the stated site is counted using sequence reads. The first methylation level of the first chromosome region is calculated by summing the corresponding numbers of methylated DNA molecules at sites within the first chromosome region. The first methylation level is the proportion of methylated DNA molecules at sites within the first chromosome region. Compare the first methylation level with a first threshold; and The classification of abnormalities in the first chromosomal region of the first tissue is determined based on the comparison. 97. The method of claim 96, wherein comparing the first methylation level with a threshold comprises: The first methylation level is standardized and the standardized first methylation level is compared with the threshold. 98. The method of claim 97, wherein the normalization refers to normalization using the second methylation level of the second chromosome region. 99. The method of claim 97, wherein the standardization refers to standardization using a percentage concentration of free DNA from the first tissue. 100. The method of claim 96, wherein calculating the first methylation level comprises: Identify multiple regions of the genome of the organism; Identify one or more sites within each of the plurality of regions; Calculate the regional methylation level for each of the plurality of regions to generate a regional methylation level, wherein the first methylation level is for a first region, and The comparison of the first methylation level with the first threshold includes: Compare the methylation level of each region with the threshold of the corresponding region; Determine a first number of regions whose regional methylation levels exceed the corresponding regional threshold; and The first number is compared with a threshold to determine the classification. 101. The method of claim 100, wherein the threshold is a percentage, and wherein comparing the first number with the threshold comprises: Before comparing with the threshold, the first number of regions is divided by the second number of regions. 102. The method of claim 101, wherein the second number of regions refers to all of the identified plurality of regions. 103. The method of claim 100, wherein the corresponding region threshold is a specified amount from a reference methylation level. 104. The method of claim 96, wherein the biological sample is derived from a female individual carrying a fetus, wherein the first tissue is derived from a fetus or placenta and the second tissue is derived from the female individual, and wherein the chromosomal abnormality is a fetal chromosomal abnormality. 105. The method of claim 104, wherein the threshold refers to the background methylation level associated with the DNA of the second tissue from the female individual. 106. The method of claim 104, wherein the threshold is determined from other female pregnant individuals carrying fetuses without chromosomal abnormalities for the first chromosomal region. 107. The method of claim 104, wherein the chromosomal abnormality refers to trisomy 21, trisomy 18, trisomy 13, Turner syndrome, or Klinefelter syndrome. 108. The method of claim 96, wherein the first tissue is derived from a tumor within the organism and the second tissue is derived from a non-malignant tissue within the organism. 109. The method of claim 96, wherein the threshold is determined by another organism that does not have cancer. 110. The method of claim 96, wherein the threshold is based on the concentration of free DNA derived from the first tissue in the biological sample. 111. The method of claim 110, wherein the threshold is based on a scaling factor corresponding to an anomaly type, wherein the anomaly type is a missing or duplicate. 112. The method of claim 96, wherein determining the location of the DNA molecule in the reference genome includes determining whether the location is within the first chromosomal region. 113. The method of claim 112, wherein determining the location of the DNA molecule in the reference genome is achieved by determining whether the DNA molecule aligns to the first chromosomal region. 114. The method of claim 96, wherein the chromosomal abnormality is a subchromosomal deletion, a subchromosomal duplication, or DiGeorge syndrome. 115. The method of any one of claims 1-89 and 91-114, wherein the biological sample is harvested by a treatment that yields free nucleic acid molecules in the biological sample. 116. The method of claim 115, wherein the biological sample is selected from plasma and serum. 117. The method of claim 115, wherein the biological sample is harvested by centrifugation. 118. A computer system comprising one or more processors and a computer-readable medium storing a plurality of instructions, which, when executed, control one or more processors of the computer system to perform the method as claimed in any one of claims 91-114. 119. A system for analyzing biological samples from an organism, said biological sample comprising a mixture of free deoxyribonucleic acid (DNA) molecules derived from normal cells and cancer-associated cells, said system comprising: A device for analyzing multiple cell-free DNA molecules from the biological sample, the analysis being for sequence reads obtained from massively parallel sequencing, and wherein the analysis of each of the multiple cell-free DNA molecules includes: To determine the location of free DNA molecules within the genome of the organism; Determine whether the free DNA molecule is methylated at one or more sites; For each of the multiple sites: A device for counting the number of free DNA molecules methylated at the site from the biological sample using sequence reads; A device for calculating a first methylation level using the sum of the respective numbers of free DNA molecules methylated at said plurality of sites, wherein said plurality of sites are located on a plurality of chromosomes; A means for comparing the first methylation level with a first threshold; and A device for determining a first category of cancer grade based on the comparison. 120. The system of claim 119, wherein the means for determining whether the free DNA molecule is methylated at one or more sites comprises: A device used for sequencing that can identify methylation. 121. The system of claim 120, wherein the means for performing sequencing that can identify methylation comprises: Apparatus for treating the free DNA molecules with sodium bisulfite; and Apparatus for sequencing the processed free DNA molecules. 122. The system of claim 121, wherein the treatment of the free DNA molecule with sodium bisulfite is part of Tet-assisted bisulfite conversion or oxidative bisulfite sequencing for the detection of 5-hydroxymethylcytosine. 123. The system of claim 120, wherein the corresponding number of free DNA molecules in each region is determined by comparing sequence reads obtained from sequencing of the recognizable methylation. 124. The system of claim 119, wherein the means for determining whether the free DNA molecule is methylated at one or more sites comprises means for digestion with methylation-sensitive restriction enzymes, methylation-specific PCR, methylation-dependent DNA precipitation, methylated DNA-binding proteins/peptides, or single-molecule sequencing without sodium bisulfite treatment. 125. The system of claim 119, further comprising: For each of the first plurality of regions of the genome: A device for determining the corresponding number of free DNA molecules from the region of the biological sample; A means for calculating the corresponding standardized value of the corresponding number of free DNA molecules from the region; and A device for comparing the corresponding standardized value with a reference value to determine whether the corresponding region shows a deficiency or an amplification; A means for determining a first amount of regions in a first plurality of regions of the genome that are identified as containing deletions or amplifications; A device for comparing the first quantity with a first threshold to determine a second category of cancer grade; and A device for determining a third category of cancer grade using the first and second categories. 126. The system of claim 125, wherein the first threshold is a percentage for determining the first plurality of regions that are missing or expanded. 127. The system of claim 125, wherein the third category is cancer-positive only when both the first category and the second category indicate cancer. 128. The system of claim 125, wherein the third category is cancer-positive when the first or second category indicates cancer. 129. The system of claim 119, wherein the first classification indicates the presence of cancer in the organism, the system further comprising: A device for identifying a type of cancer associated with an organism by comparing the first methylation level with a corresponding value determined from other organisms, wherein at least two of the other organisms are identified as having different types of cancer. 130. The system of claim 125, wherein the first classification indicates the presence of cancer in the organism, the system further comprising: A device for identifying a type of cancer associated with an organism by comparing the first methylation level with a corresponding value determined from other organisms, wherein at least two of the other organisms are identified as having different types of cancer. 131. The system of claim 130, wherein the third classification indicates the presence of cancer in the organism, the system further comprising: A means for identifying the type of cancer associated with the organism by comparing the first quantity in the region with a corresponding value determined from the other organism. 132. The system of claim 125, wherein the means for calculating the first methylation level comprises: Device for identifying a second plurality of regions of the genome; A means for identifying one or more sites within each of the second plurality of regions; A means for calculating the regional methylation level of each of the second plurality of regions, wherein the first methylation level is for a first region. The system further includes: A means for comparing each of the methylation levels of the regions with a corresponding region threshold, comprising comparing the first methylation level with the first threshold; A means for determining a second quantity of regions whose methylation levels exceed a corresponding regional threshold; and A means for comparing the second quantity of a region in the second plurality of regions with a second threshold to determine the first classification. 133. The system of claim 119, wherein the means for calculating the first methylation level comprises: Device for identifying a second plurality of regions of the genome; A means for identifying one or more sites within each of the second plurality of regions; A means for calculating the regional methylation level of each of the second plurality of regions, wherein the first methylation level is for a first region. The system further includes: A means for comparing each of the methylation levels of the regions with a corresponding region threshold, comprising comparing the first methylation level with the first threshold; A means for determining a second quantity of regions whose methylation levels exceed a corresponding regional threshold; and A means for comparing the second quantity of a region in the second plurality of regions with a second threshold to determine the first classification. 134. The system of claim 133, wherein the regions where the methylation level exceeds the corresponding region threshold correspond to a first group of regions, the system further comprising: A device for comparing the regional methylation level of the first set of regions with the corresponding regional methylation level of other organisms regarding the first set of regions, said other organisms having at least two of the following: a first type of cancer, no cancer, and a second type of cancer; and A device for determining, based on the comparison, whether the organism has the first type of cancer, does not have cancer, or has the second type of cancer. 135. The system of claim 134, further comprising: A device for clustering other organisms based on the methylation levels of corresponding regions in a first group of regions, wherein two of the groups correspond to either: the first type of cancer, the absence of cancer, and the second type of cancer. The comparison of regional methylation levels in the second plurality of regions is used to determine which category the organism belongs to. 136. The system of claim 135, wherein the clustering of the other organisms uses the regional methylation level of the organism. 137. The system of claim 135, wherein the class includes a first class corresponding to the first type of cancer, a second class corresponding to the second type of cancer, and a third class corresponding to the absence of cancer. 138. The system of claim 135, wherein the clustering of the other organisms is further based on corresponding normalized values of a second set of regions of the other organisms, wherein the second set of regions corresponds to regions identified as containing deletions or amplifications, and wherein the corresponding normalized value of the region is determined by the corresponding number of free DNA molecules from the region, the system further comprising: For each of the regions in the second group: A device for determining the corresponding number of free DNA molecules from said region; and A means for calculating a corresponding normalized value from the corresponding number of free DNA molecules from said region; and A device for comparing the corresponding standardized values of the second group of regions of the organism with the corresponding standardized values of the other organisms to determine which category the organism belongs to. 139. The system of claim 138, wherein the clustering of the other organisms is further based on the corresponding methylation density of highly methylated CpG islands, the system further comprising: For each of the hypermethylated CpG islands: Apparatus for determining the corresponding methylation density, and An apparatus for comparing the methylation density of the hypermethylated CpG islands of the organism with the methylation density of other organisms to determine which category the organism belongs to. 140. The system of claim 133, further comprising: For each of the second plurality of regions: A device for calculating the difference between the methylation level of the region and the corresponding threshold of the region; and A device for calculating the corresponding probability corresponding to the corresponding difference; The means for determining the second quantity of the region includes: A device for calculating a cumulative score including the corresponding probabilities. 141. The system of claim 140, wherein the means for calculating the cumulative score comprises: A means for obtaining a corresponding logarithmic result by utilizing the logarithm of the corresponding probability of each region in the second plurality of regions; and A means for calculating the sum including the corresponding logarithmic results. 142. The system of claim 141, wherein the cumulative score is the negative of the sum of the corresponding logarithmic results. 143. The system of claim 140, wherein the corresponding differences in each of the second plurality of regions are normalized by the standard deviation of the corresponding region threshold. 144. The system of claim 140, wherein the corresponding probability corresponds to the probability of the corresponding difference according to a statistical distribution. 145. The system of claim 140, wherein the second threshold corresponds to the highest cumulative score of a reference group of samples from other organisms. 146. The system of claim 133, further comprising: For each of the first plurality of regions: A means for calculating the corresponding difference between the corresponding standardized value and the reference value; and A device for calculating the corresponding probability corresponding to the corresponding difference; The means for determining the first quantity of the region includes: A means for calculating a first sum including the corresponding probabilities. 147. The system of claim 133, wherein the corresponding region threshold is a specified amount derived from a reference methylation level. 148. The system of claim 133, wherein the second threshold is a percentage, and wherein the second quantity in the comparison region is compared with the second threshold, comprising: A means for dividing the second quantity of a region by the second number of the second plurality of regions before comparing it with the second threshold. 149. The system of claim 148, wherein the second number corresponds to all of the second plurality of regions. 150. The system of claim 133, wherein the first plurality of regions are the same as the second plurality of regions, and wherein the threshold of the corresponding region depends on whether the corresponding region shows a deficiency or an amplification. 151. The system of claim 150, wherein, compared to when no amplification is displayed, one of the threshold values of the corresponding region when amplification is displayed has a larger value, and wherein, compared to when no deficiency is displayed, the second of the threshold values of the corresponding region when deficiency is displayed has a smaller value. 152. The system of claim 151, wherein the corresponding region threshold tests the hypomethylation of the second plurality of regions, wherein the third of the corresponding region thresholds has a larger negative value when the corresponding region shows amplification compared to when no amplification is shown, and wherein the fourth of the corresponding region thresholds has a smaller negative value when the corresponding region shows deletion compared to when no deletion is shown. 153. The system of claim 133, wherein the biological sample is collected prior to treatment, further comprising: For another biological sample collected after treatment, the system as described in claim 133 is used to obtain: Determine the first subsequent measure to show the region of absence or amplification; and A subsequent second quantity is determined for regions where the methylation level exceeds the corresponding regional threshold; The first quantity is compared with the subsequent first quantity, and the second quantity is compared with the subsequent second quantity to determine the prognosis of the organism. 154. The system of claim 153, wherein comparing the first quantity with the subsequent first quantity and comparing the second quantity with the subsequent second quantity to determine the prognosis of the organism comprises: A means for determining a first difference between the first quantity and the subsequent first quantity; A means for comparing the first difference with one or more first difference thresholds; A means for determining a second difference between the second quantity and the subsequent second quantity; and A means for comparing the second difference with one or more second difference thresholds. 155. The system of claim 154, wherein when the first difference is below one of the first difference thresholds, the prognosis is predicted to be worse compared to when the first difference exceeds one of the first difference thresholds, and wherein when the second difference is below one of the second difference thresholds, the prognosis is predicted to be worse compared to when the second difference exceeds one of the second difference thresholds. 156. The system of claim 155, wherein one of the first difference thresholds and one of the second difference thresholds are zero. 157. The system of claim 153, wherein the treatment is immunotherapy, surgery, radiation therapy, chemotherapy, antibody-based therapy, epigenetic therapy, or targeted therapy. 158. The system of claim 119, wherein the first threshold is a specified distance from a reference methylation level established by a biological sample obtained from a healthy organism. 159. The system of claim 158, wherein the specified distance is a specified number of standard deviations relative to the reference methylation level. 160. The system of claim 119, wherein the first threshold is established by a reference methylation level determined from a previous biological sample of the organism obtained prior to the biological sample test. 161. The system of claim 119, wherein the means for comparing the first methylation level with the first threshold comprises: A means for determining the difference between the first methylation level and a reference methylation level; and A means for comparing the difference with a value corresponding to the first threshold. 162. The system of claim 119, further comprising: A device for determining the percentage concentration of cell-free tumor DNA in the biological sample; A device for calculating the first threshold based on the percentage concentration. 163. The system of claim 119, further comprising: A device for determining whether the percentage concentration of cell-free tumor DNA in the biological sample exceeds a minimum value; and An apparatus for labeling a biological sample if the percentage concentration does not exceed the minimum value. 164. The system of claim 163, wherein the minimum value is determined based on the expected difference between the tumor methylation level and the reference methylation level. 165. The system of claim 119, further comprising: A device for measuring the size of free DNA molecules located at the plurality of sites; and An apparatus for normalizing the first methylation level based on the measured size of the free DNA molecule before comparing the first methylation level with the first threshold. 166. The system of claim 165, wherein the means for normalizing the first methylation level based on the measurement size comprises: A device for selecting free DNA molecules of a first size; A means for calculating the first methylation level using selected free DNA molecules, wherein the first threshold corresponds to the first size. 167. The system of claim 166, wherein the first dimension is a length range. 168. The system of claim 166, wherein the free DNA molecules are selected based on size-dependent physical separation. 169. The system of claim 166, wherein the means for selecting free DNA molecules having a first size comprises: Apparatus for performing paired-end massive parallel sequencing on the plurality of free DNA molecules to obtain paired sequences of each of the plurality of free DNA molecules; A device for determining the size of free DNA molecules by comparing the paired sequences of the plurality of free DNA molecules with a reference genome; and A device for selecting free DNA molecules having the first size. 170. The system of claim 165, wherein the means for normalizing the first methylation level based on the size of the measurement comprises: Apparatus for obtaining a functional relationship between size and methylation level; and A device for normalizing the first methylation level using the functional relationship. 171. The system of claim 170, wherein the functional relationship provides correction coefficients corresponding to the respective dimensions. 172. The system of claim 171, further comprising: A device for calculating the average size of a free DNA molecule corresponding to the first methylation level; and A means for multiplying the first methylation level by the corresponding correction coefficient. 173. The system of claim 171, further comprising: For each of the plurality of sites: For each of the free DNA molecules located at the said site: A device for obtaining the corresponding size of the free DNA molecule at the said site; and A device for normalizing the contribution of the free DNA molecule to the corresponding number of free DNA molecules methylated at the site using the correction coefficient corresponding to the corresponding size. 174. The system of claim 119, wherein the plurality of sites comprises CpG sites, wherein the CpG sites are organized into a plurality of CpG islands, each CpG island comprising a plurality of CpG sites, and wherein the first methylation level corresponds to a first CpG island of the plurality of CpG islands. 175. The system of claim 174, wherein each of the CpG islands in a reference population of samples from other organisms has an average methylation density of less than a first percentage, and wherein each of the CpG islands in the reference population has an average methylation density coefficient of variation of less than a second percentage, and wherein for each CpG island, each methylation density is determined using the total number of methylated and unmethylated DNA molecules spanning the plurality of CpG sites. 176. The system of claim 174, further comprising: For each of the CpG islands: A device for determining whether a CpG island is highly methylated relative to a reference group of samples from other organisms by comparing the methylation level of the CpG island with a corresponding threshold; For each of the hypermethylated CpG islands: Apparatus for determining the corresponding methylation density, A means for calculating the cumulative fraction from the corresponding methylation density; and A means for comparing the cumulative score with a cumulative threshold to determine the first classification. 177. The system of claim 176, wherein the means for calculating the cumulative fraction from the corresponding methylation density comprises: For each of the hypermethylated CpG islands: A device for calculating the corresponding difference between the corresponding methylation density and the reference density; and A means for calculating the corresponding probability corresponding to the corresponding difference; and A means for determining the cumulative score using the corresponding probability. 178. The system of claim 177, wherein the cumulative score is determined by: For each of the hypermethylated CpG islands: The corresponding logarithmic result is obtained by using the logarithm of the corresponding probability; and The calculation includes the sum of the corresponding logarithmic results, and the cumulative score is the negative of the sum. 179. The system of claim 177, wherein each corresponding difference is standardized using a standard deviation associated with the reference density. 180. The system of claim 176, wherein the cumulative threshold corresponds to the highest cumulative score from the reference group. 181. The system of claim 176, wherein the means for determining whether the first CpG island is hypermethylated comprises: A device for comparing the first methylation level with the first threshold and the third threshold. The first threshold corresponds to the average methylation density of the reference population plus a specified percentage, and the third threshold corresponds to a specified number of standard deviations plus the average methylation density of the reference population groups. 182. The system of claim 181, wherein the specified percentage is 2%. 183. The system of claim 181, wherein the specified number of standard deviations is three. 184. The system of claim 119, further comprising: For each of the first plurality of regions of the genome: A device for determining the corresponding number of free DNA molecules from said region; A means for calculating a corresponding standardized value from the corresponding number; and A device for comparing the corresponding standardized value with a reference value to determine whether the region shows a missing or amplified feature; A means for determining a first set of regions, the first set of regions being determined to have one of the following: Deletion, amplification, or normal presentation, wherein the first methylation level corresponds to the first group of regions; A means for determining a second set of regions, the second set of regions being determined to have another of the following: deletion, amplification, or normal presentation; and A device for calculating the second methylation level based on the corresponding number of free DNA molecules methylated at sites in the second group of regions. The means for comparing the first methylation level with the first threshold includes: A device for calculating a parameter between the first methylation level and the second methylation level; and A means for comparing the parameter with the first threshold. 185. The system of claim 184, wherein the first methylation level is a statistical value of the regional methylation level calculated for each region of the first group of regions, and wherein the second methylation level is a statistical value of the regional methylation level calculated for each region of the second group of regions. 186. The system of claim 185, wherein the statistic is determined using a Student’s t-test, an analysis of variance (ANOVA) test, or a Kruskal-Wallis test. 187. The system of claim 184, wherein the parameter includes the ratio or difference between the first methylation level and the second methylation level. 188. The system of claim 187, wherein the means for calculating the parameter includes means for applying a probability distribution to the ratio or the difference. 189. A system for determining a first methylation pattern from a biological sample of an organism, said biological sample comprising free DNA containing a mixture of free DNA molecules derived from a first tissue and a second tissue, said system comprising: Apparatus for obtaining a second methylation pattern corresponding to DNA molecules of the second tissue, the second methylation pattern providing a second methylation density at each of a plurality of loci in the genome of the organism, the second methylation density at a particular locus being the proportion of methylated DNA molecules of the second tissue at that locus; A device for determining the free methylation pattern from the free DNA molecules of the mixture, the free methylation pattern providing a mixed methylation density at each of the plurality of loci, the mixed methylation density at each locus being the proportion of methylated DNA molecules of the mixture at that locus; A device for determining the percentage of free DNA molecules from the first tissue in the mixture; and Apparatus for determining the first methylation type of the first tissue by: For each of the plurality of loci: An apparatus for calculating a difference parameter, the difference parameter including the difference between the second methylation density of the second methylation type and the mixed methylation density of the free methylation type, the difference being measured by the percentage of the free DNA molecules from the first tissue in the mixture. 190. The system of claim 189, further comprising: A means for determining a first methylation level from the first methylation morphology; and A device for comparing the first methylation level with a first threshold to determine the classification of cancer grade. 191. The system of claim 189, further comprising: A device for changing the first methylation type to obtain a corrected first methylation type. 192. The system of claim 191, wherein the transformation is a linear transformation. 193. The system of claim 189, wherein the differential parameter D of the locus is defined as follows: mbc represents the second methylation density of the second methylation pattern at the locus, mp represents the mixed methylation density of the free methylation pattern at the locus, f is the percentage concentration of free DNA molecules from the first tissue in the biological sample, and CN represents the copy number at the locus. 194. The system of claim 193, wherein CN is one for the copy number of the first tissue that is diploid at the locus. 195. The system of claim 193, further comprising: A device for identifying regions where D exceeds a threshold. 196. The system of claim 189, further comprising: A device for identifying the plurality of loci by selecting loci having one or more of the following criteria: GC content exceeding 50%; The second methylation density of the second methylation type is below the first threshold or exceeds the second threshold; and The region defined by the locus contains at least five CpG sites. 197. The system of claim 189, wherein the biological sample is derived from a female individual carrying a fetus, and wherein the first tissue is derived from the fetus or placenta and the second tissue is derived from the female individual. 198. The system of claim 197, further comprising: A means for determining a first methylation level from the first methylation morphology; and A device for comparing the first methylation level with a first threshold to determine the classification of a medical condition. 199. The system of claim 198, wherein the medical condition is preeclampsia or a chromosomal abnormality in the first tissue. 200. The system of claim 189, wherein the first tissue is derived from a tumor within the organism and the second tissue is derived from non-malignant tissue within the organism. 201. The system of claim 189, wherein the biological sample further comprises DNA molecules from the second tissue, and wherein obtaining the second methylation form comprises: Apparatus for analyzing the DNA molecules from the second tissue to determine the second methylation pattern. 202. The system of claim 189, wherein the obtained second methylation type corresponds to the average methylation type obtained from a set of control samples. 203. The system of claim 189, wherein determining the mixed methylation density of the methylation patterns of the free DNA at the first locus of the plurality of loci comprises: A device for determining a first number of the free DNA molecules from the first locus; A means for determining whether each of the first number of DNA molecules is methylated at one or more sites to obtain a second number of methylated DNA molecules; and A device for calculating the mixed methylation density from the first number and the second number. 204. The system of claim 203, wherein the one or more sites are CpG sites. 205. The system of claim 189, wherein the means for determining the methylation pattern of the free DNA comprises: Apparatus for treating the free DNA molecules with sodium bisulfite; and Apparatus for sequencing the processed DNA molecules. 206. The system of claim 205, wherein the treatment of the free DNA molecule is part of a Tet-assisted bisulfite conversion or includes treatment of the free DNA molecule with potassium perruthenate (KRuO4). 207. The system of claim 189, wherein the means for determining the free methylation type comprises: An apparatus for performing single-molecule sequencing on the DNA, wherein the single-molecule sequencing includes determining whether at least one site of the DNA molecule is methylated. 208. A system for determining a first methylation pattern from a biological sample of an organism, said biological sample comprising free DNA containing a mixture of free DNA molecules derived from a first tissue and a second tissue, said system comprising: A device for analyzing multiple DNA molecules from said biological sample, wherein the DNA molecules analyzed from said multiple DNA molecules include: A device for determining the location of the DNA molecule in the genome of the organism; A device for determining the genotype of the DNA molecule; and A device for determining whether the DNA molecule is methylated at one or more sites; A device for identifying multiple first loci, wherein the first allele and the corresponding second allele corresponding to the first genome of the first tissue are heterozygous and the first allele corresponding to the second genome of the second tissue is homozygous; For each of the first loci: For each of the one or more loci associated with the said locus: A device for determining the number of DNA molecules methylated at the site and corresponding to the corresponding second allele at the locus. A means for calculating methylation density using the number of DNA molecules methylated at one or more sites at the locus and corresponding to the respective second allele of the locus; and An apparatus for generating the first methylation pattern of the first tissue from the methylation density of the first locus, wherein the first methylation density at each locus is the proportion of methylated DNA molecules of the first tissue at the locus. 209. The system of claim 208, wherein the biological sample is derived from a female individual carrying a fetus, and wherein the first tissue is derived from the fetus or placenta and the second tissue corresponds to the female individual. 210. The system of claim 208, wherein the site corresponds to a CpG site. 211. The system of claim 208, wherein each locus includes at least one CpG site. 212. The system of claim 208, wherein each locus is adjacent to at least one CpG site. 213. A system for detecting chromosomal abnormalities from a biological sample of an organism, said biological sample comprising cell-free DNA containing a mixture of cell-free DNA molecules derived from a first tissue and a second tissue, said system comprising: An apparatus for analyzing multiple DNA molecules from said biological sample, said analysis being for sequence reads obtained from massively parallel sequencing, wherein the DNA molecules analyzed comprise: A device for determining the location of the DNA molecule in a reference genome; and A device for determining whether the DNA molecule is methylated at one or more sites; For each of the multiple sites: A device for determining the corresponding number of DNA molecules methylated at said site using sequence reads; A device for calculating a first methylation level of a first chromosome region by summing the corresponding numbers of methylated DNA molecules at sites within the first chromosome region, wherein the first methylation level is the proportion of methylated DNA molecules at sites within the first chromosome region. A means for comparing the first methylation level with a first threshold; and A device for classifying abnormalities in the first chromosomal region of the first tissue based on the comparison. 214. The system of claim 213, wherein the means for comparing the first methylation level with a threshold comprises: A device for standardizing the first methylation level and comparing the standardized first methylation level with the threshold. 215. The system of claim 214, wherein the normalization refers to normalization using the second methylation level of the second chromosome region. 216. The system of claim 214, wherein the normalization refers to normalization using a percentage concentration of free DNA from the first tissue. 217. The system of claim 213, wherein the means for calculating the first methylation level comprises: Device for identifying multiple regions of the genome of the organism; A means for identifying one or more sites within each of the plurality of regions; A means for calculating the regional methylation level of each of the plurality of regions to generate a regional methylation level, wherein the first methylation level is for a first region, and The means for comparing the first methylation level with a first threshold includes: A means for comparing the methylation level of each of the regions with a threshold for the corresponding region; A means for determining a first number of regions whose regional methylation levels exceed the corresponding regional threshold; and A means for determining the classification by comparing the first number with a threshold. 218. The system of claim 217, wherein the threshold is a percentage, and wherein the means for comparing the first number with the threshold comprises: A means for dividing the first number of regions by the second number of regions before comparing with the threshold. 219. The system of claim 218, wherein the second number of regions refers to all of the identified plurality of regions. 220. The system of claim 217, wherein the corresponding region threshold is a specified amount from a reference methylation level. 221. The system of claim 213, wherein the biological sample is derived from a female individual carrying a fetus, wherein the first tissue is derived from the fetus or placenta and the second tissue is derived from the female individual, and wherein the chromosomal abnormality is a fetal chromosomal abnormality. 222. The system of claim 221, wherein the threshold refers to the background methylation level associated with the DNA of the second tissue from the female individual. 223. The system of claim 221, wherein the threshold is determined from other female pregnant individuals carrying fetuses without chromosomal abnormalities for the first chromosomal region. 224. The system of claim 221, wherein the chromosomal abnormality refers to trisomy 21, trisomy 18, trisomy 13, Turner syndrome, or Klinefelter syndrome. 225. The system of claim 213, wherein the first tissue is derived from a tumor within the organism and the second tissue is derived from non-malignant tissue within the organism. 226. The system of claim 213, wherein the threshold is determined by another organism that does not have cancer. 227. The system of claim 213, wherein the threshold is based on the concentration of free DNA derived from the first tissue in the biological sample. 228. The system of claim 227, wherein the threshold is based on a scaling factor corresponding to an anomaly type, wherein the anomaly type is a missing or duplicate. 229. The system of claim 213, wherein the means for determining the location of the DNA molecule in a reference genome includes means for determining whether the location is within the first chromosomal region. 230. The system of claim 229, wherein the means for determining the location of the DNA molecule in the reference genome is implemented by means for determining whether the DNA molecule aligns to the first chromosomal region. 231. The system of claim 213, wherein the chromosomal abnormality is a subchromosomal deletion, a subchromosomal duplication, or DiGeorge syndrome. 232. The system of any one of claims 119-231, wherein the biological sample is harvested by a process for obtaining free nucleic acid molecules in the biological sample. 233. The system of claim 232, wherein the biological sample is selected from plasma and serum. 234. The system of claim 232, wherein the biological sample is harvested by centrifugation.