KR20230157204A - Method for detecting aneuploidy of fetus based on synthetic positive data and synthetic negative data - Google Patents

Abstract

Provided is a method for detecting chromosomal aneuploidy in a fetus based on virtual normal data and virtual positive data, and a computer-readable medium recording a program applied to perform the same. According to this, chromosomal aneuploidy in the fetus can be non-invasively diagnosed prenatally with excellent sensitivity and specificity.

Description

Method for detecting aneuploidy of fetus based on synthetic positive data and synthetic negative data}

A method for detecting chromosomal aneuploidy in a fetus based on virtual positive data and virtual negative data, that is, virtual chromosomal aneuploidy data and virtual normal data, is provided.

Prenatal diagnosis refers to diagnosing the presence or absence of a disease in the fetus before the fetus is born. Prenatal diagnosis is largely divided into invasive and non-invasive diagnostic methods. Invasive diagnostic methods include, for example, chorionic screen, amniocentesis, and umbilical cord puncture. Invasive diagnostic methods have the potential to cause miscarriage, disease, or deformity by causing shock to the fetus during the test process, so non-invasive diagnostic methods are being developed.

Recently, it has been demonstrated that noninvasive diagnosis of fetal chromosomal aneuploidy is feasible by massively parallel sequencing of DNA molecules in the plasma of pregnant women. Fetal DNA can be detected in maternal plasma and serum from the 7th week of pregnancy, and the amount of fetal DNA in maternal blood increases with the period of pregnancy. When performing massively parallel sequencing of fetal DNA, the threshold for distinguishing a normal fetus from a chromosomal aneuploidy fetus is unclear, resulting in low sensitivity and specificity for chromosomal aneuploidy detection.

Therefore, there is a need to develop a method that can clearly distinguish between normal fetuses and chromosomal aneuploidy fetuses and increase the sensitivity and specificity of chromosomal aneuploidy detection.

KR 1739535 B

A method for detecting chromosomal aneuploidy in a fetus is provided.

A computer-readable medium recording a program applied to perform the method is provided.

In one aspect, as a method for detecting chromosomal aneuploidy in a fetus,

Obtaining sequence information (Read) of a plurality of nucleic acid fragments obtained from a biological sample of a woman pregnant with a normal fetus;

Mapping the obtained sequence information to a human reference genome and assigning the sequence information of the nucleic acid fragment to a chromosome;

Based on the sequence information of the nucleic acid fragment specified in the chromosome, read count (RC), fetal DNA fraction (FF), and GC content of the sequence information of the nucleic acid fragment in the target chromosome are calculated in a normal fetus. steps;

A step of producing virtual normal data from normal data based on the number of reads in the target chromosome and fetal DNA fraction in the normal fetus,

The normal data is the number of reads in the target chromosome of a normal fetus;

Generating virtual positive data from the virtual normal data,

The positive data is the number of reads in the target chromosome of a fetus with chromosomal aneuploidy;

Establishing a chromosomal aneuploidy discrimination model from the produced virtual normal data and virtual positive data; and

A method including the step of detecting chromosomal aneuploidy in a test sample using the chromosomal aneuploidy determination model is provided.

The term “chromosomal aneuploidy” refers to a condition in which the number of chromosomes per cell in a cell, individual, or lineage is not an integer multiple of the base number, and one to several more or fewer chromosomes per integer multiple, that is, a genome with an incomplete structure. refers to a state that includes In the case of diploidy, when two homologous chromosomes of a pair are missing, it is called chromosome 0, when one pair is missing and only the other chromosome is present, it is called chromosome 1, and in addition to the pair of homologous chromosomes, another extra chromosome is present. This case is called trisomy (T).

The chromosomal aneuploidy may be trisomy 13, trisomy 18, trisomy 21, XO, XXX, XXY, XYY, or a combination thereof. Abnormalities of chromosome 13 (trisomy 13) are associated with Patau syndrome. Abnormalities of chromosome 18 (trisomy 18) are associated with Edwards syndrome. Abnormalities of chromosome 21 (trisomy 21) are associated with Down syndrome. Monosomy X (XO, i.e. absence of one chromosome X) is associated with Turner syndrome. XXY is a condition in which human males have an extra chromosome X and is associated with Klinefelter syndrome.

The method includes obtaining sequence information (Read) of a plurality of nucleic acid fragments obtained from a biological sample of a woman carrying a normal fetus.

The pregnant woman may be a woman pregnant with a single fetus or twin fetuses.

The biological sample may be blood, plasma, serum, urine, saliva, mucosal secretions, sputum, feces, tears, or a combination thereof. The biological sample is, for example, peripheral blood plasma. The biological sample may include fetal nucleic acid. The fetal nucleic acid may be cell-free DNA (cfDNA). The fetal nucleic acid may be isolated DNA.

The step of obtaining sequence information of a plurality of nucleic acid fragments obtained from the biological sample of the pregnant woman may include the step of isolating the nucleic acids from the biological sample.

The method of isolating nucleic acids from the biological sample can be performed by methods known to those skilled in the art. The length of the isolated nucleic acid fragment is about 10 bp (base pairs) to about 2000 bp, about 15 bp to about 1500 bp, about 20 bp to about 1000 bp, about 20 bp to about 500 bp, about 20 bp to about 200 bp. , or may be from about 20 bp to about 100 bp.

The step of obtaining sequence information of a plurality of nucleic acid fragments from the biological sample obtained from the pregnant woman may include performing massive parallel sequencing on the isolated nucleic acids.

The term “massive parallel sequencing” may be used interchangeably with next-generation sequencing (NGS) or second-generation sequencing. Massively parallel sequencing refers to a technique for simultaneously sequencing millions of fragments of nucleic acids. Massively parallel sequencing is available, for example, on the 454 platform (Roche), GS FLX Titanium, Illumina MiSeq, Illumina HiSeq, Illumina Genome Analyzer, Solexa platform, SOLiD System (Applied Biosystems), Ion Proton (Life Technologies), Complete Genomics, Helicos Biosciences. It can be performed in parallel by Heliscope, Pacific Biosciences' single molecule real-time (SMRT™) technology, or a combination of these.

The method may further include preparing a nucleic acid library to perform massively parallel sequencing.

The nucleic acid library can be prepared according to massively parallel sequencing. Nucleic acid libraries can be constructed following the manufacturer's instructions that provide massively parallel sequencing.

The sequence information of the obtained nucleic acid fragment may also be called a read.

The method includes mapping the obtained sequence information to a human reference genome and assigning the sequence information of the nucleic acid fragment to a chromosome.

The human reference genome may be hg18 or hg19. Sequence information that maps to only one genomic location in the human reference genome can be designated as unique sequence information. Based on the designated unique sequence number, the sequence information of the nucleic acid fragment can be assigned to the location of the chromosome. The chromosomal location may be a contiguous range on a chromosome having a length of at least about 5 kb, about 10 kb, about 20 kb, about 50 kb, about 100 kb, about 1000 kb, or 2000 kb. The chromosomal location may be a single chromosome.

After the step of assigning the sequence information of the nucleic acid fragment to the chromosome, the step of checking the thickness distribution of the sequence information of the nucleic acid fragment assigned to the chromosome for each section and excluding a section with low reliability of the sequence information from the analysis target. You can. The section may be set in units of about 5 kb to about 50 kb. For example, the section may be set to about 10 kb to about 40 kb, about 15 kb to about 30 kb, or about 20 kb to about 25 kb. By setting the above section, filtering can be done using the GC content of the base sequence. In addition, by setting the section, it is possible to form a group of the depth and GC content of the sequence information of the nucleic acid fragment specified in the chromosome, and statistical analysis may be possible.

The step of excluding the section with low confidence in the sequence information from the analysis target includes removing mismatch portions, removing sequence information assigned to multiple regions, removing redundant sequence information, or any of these. May include combinations. In order to exclude sections with low confidence in sequence information from analysis, quality filtering, trimming, perfect match, removal of sequences assigned to multiple locations, and PCR duplicate sequences Removal of information (PCR duplicated reads) or a combination of these can be performed. The quality filtering is a process of extracting sequence information with high quality relative to the quality of each base sequence obtained during the sequencing process. The trimming is a process of removing parts of poor quality because the quality of the latter part of the base sequence is low due to the nature of the sequencing device. For example, nucleic acid fragments can be trimmed to a size of at least about 50 bp, greater than about 50 bp, or greater than about 100 bp. For example, the quality value of the nucleic acid fragment may be 20 or higher, 30 or higher, 40 or higher, or 50 or higher. The perfect match selects only base sequences that perfectly match when mapping to the human reference genome. Since sequences assigned to multiple locations are likely to be repetitive sequence regions, sequences assigned to multiple locations (multi) can be removed from the obtained sequence information. Removing PCR duplicate sequence information means removing parts that have been amplified more due to errors during the sequencing process. In addition, for statistical analysis, a group with a certain degree of even deviation must be selected to obtain meaningful results. The part without thickness is usually the N-region of the chromosome and can be removed from the analysis target.

In this method, based on the sequence information of the nucleic acid fragment assigned to the chromosome, the total number assigned to chromosome 1 is expressed as C1RC (Chromosome 1 Read Count), and in the case of chromosome 2 assigned, it is expressed as C2RC (Chromosome 2 Read Count). You can. Therefore, the total number assigned to the ith chromosome can be expressed as CiRC (Chromosome i Read Count). Additionally, in the case of sex chromosomes, the total number assigned to chromosome X can be expressed as CXRC (Chromosome The sum of the total allocated numbers of autosomes 1 to 22 is denoted as ATRC (Autosomal Total Read Count), and the sum of the total allocated numbers of all autosomes and sex chromosomes is denoted as TRC (Total Read Count). .

The method is based on the sequence information of the nucleic acid fragment specified in the chromosome, read count (RC), fetal DNA fraction (FF), and GC content of the sequence information of the nucleic acid fragment in the target chromosome in a normal fetus. It includes the step of calculating .

The target chromosome may be selected from the group consisting of chromosomes 1 to 22, chromosome X, and chromosome Y. The target chromosome may be selected from the group consisting of chromosome 13, chromosome 18, and chromosome 21.

The term “Read Count (RC)” or “read number” refers to the number of reads assigned to the position of one gene.

The term “Fetal Fraction (FF)” or “fraction of fetal nucleic acids” refers to the amount or percentage (%) of fetal nucleic acids among nucleic acids isolated from a biological sample of a pregnant woman. The fetal fraction may be a concentration, relative ratio, or absolute amount of fetal nucleic acids. The fetal nucleic acid may be a nucleic acid derived from placental trophoblasts.

The term “GC content” refers to the ratio (%) of guanine (G) and cytosine (C) among the bases that make up DNA. The GC content can be calculated from the formula GC content=(G+C)/(A+T+G+C).

The method includes generating virtual normal data from normal data based on the number of reads in the target chromosome and fetal DNA fraction in the normal fetus.

The normal data may be the number of reads in the target chromosome of a normal fetus. The normal data may also be called voice data. The normal sample may be a normal male fetus sample or a normal female fetus sample.

The virtual normal data may be the number of reads in the target chromosome obtained by randomly combining and dividing normal data.

The step of generating virtual normal data from the normal data may be a step of randomly selecting two or more samples from among the normal samples, combining the selected fetal samples, and randomly dividing them again to generate virtual normal data. In order to produce virtual normal data from the normal data, two or more samples from among the normal samples can be randomly selected and the selected fetal samples can be combined to generate the entire sample. At this time, since each sample is a collection of leads, new virtual normal data can be generated by randomly selecting a desired amount of leads from the entire sample. The fetal DNA fraction and GC content of the hypothetical normal data can be determined as follows. If A is the number of reads in sample A, B is the number of reads in sample B, the fetal DNA fraction of sample A is Aff, and the fetal DNA fraction of sample B is denoted by Bff, the fetal DNA fraction of the entire sample is (A × Aff + B GC content can be recalculated from new virtual normal data.

The step of producing virtual normal data from the normal data includes distinguishing between normal male fetus samples and normal female fetus samples; and randomly select two or more samples from normal male fetus samples, combine the selected male fetus samples and randomly divide again to generate male fetus virtual normal data, and/or randomly select two or more samples from normal female fetus samples and select It may include combining female fetus samples and randomly dividing them again to generate virtual normal data for female fetuses.

The method includes generating virtual positive data from the virtual normal data.

The virtual positive data is the number of reads on the target chromosome of a fetus with chromosomal aneuploidy, and may be the number of virtual reads obtained from virtual normal data.

The step of producing the virtual positive data may use CiRC (chromosome i read count) (i=1 to 22) and FF to detect autosomal aneuploidy. For example, for chromosome 13, use C13RC and FF.

The step of generating virtual positive data from the virtual normal data can be determined by the formula TiRC = CiRC + CiRC × FF/2 when the target chromosome is an autosome. In the above formula, TiRC is the number of reads on target chromosome i in an aneuploidy fetus, CiRC is the number of reads on target chromosome i in a normal fetus, and FF is the fetal DNA fraction.

The step of generating virtual positive data from the virtual normal data can be determined by the equation CXXXRC = CXXRC + CXXRC × FF/2 when the target chromosome is the XXX chromosome. In the above formula, CXXXRC is the number of reads on target chromosome XXX in a chromosomal aneuploidy fetus, CXXRC is the number of reads on target chromosome XX in a normal female fetus, and FF is the fetal DNA fraction.

The step of generating virtual positive data from the virtual normal data can be determined by the formula CXRC = CXYRC - CYRC + CyRC when the target chromosome is an XO chromosome. In the above formula, CXRC is the number of reads on target chromosome XO in a chromosomal aneuploidy fetus, CXYRC is the number of reads on target chromosome This is the number of reads incorrectly assigned to target chromosome Y. The CXRC, CXYRC, CYRC, and CyRC may be virtual data.

The step of generating virtual positive data from the virtual normal data can be determined by the equation CXYYRC = CXYRC + CYRC - CyRC when the target chromosome is an XYY chromosome. In the above formula, CXYYRC is the number of reads on target chromosome XYY in a chromosomal aneuploidy fetus, CXYRC is the number of reads on target chromosome This is the number of reads incorrectly assigned to target chromosome Y. The CXYYRC, CXYRC, CYRC, and CyRC may be virtual data.

The CyRC calculates a residual from CyRC = β ₀ + β ₁ TRC + ε (Equation 1) and CyRC' = β ₀ + β ₁ TRC (Equation 2); And it can be calculated from the step of calculating CyRC from the residual. In Equation 1, TRC is the number of reads of the entire chromosome, β ₀ is the intercept, β ₁ is the coefficient between TRC and CyRC, and ε is the residual. In Equation 2, CyRC' is the predicted value of CyRC, TRC is the number of reads of the entire chromosome, β ₀ is the fragment, and β ₁ is the coefficient between TRC and CyRC'. Female fetal samples theoretically have no leads assigned to chromosome Y. However, some reads are incorrectly assigned to chromosome Y, and the number of reads tends to increase in proportion to the size of the TRC. Therefore, Equations 1 and 2 above can be modeled using female fetus (XX) samples. The model established in this way can be approximately applied to male fetus (XY) samples that differ only in sex chromosomes from female fetus samples. This is because the number of reads incorrectly assigned to chromosome Y in male fetus samples is similar to that in female fetus samples in proportion to the size of the TRC. In other words, the TRC of the female fetus (XX) sample can be regarded as the TRC of the male fetus (XY) sample. The distribution of the residuals can be assumed to follow a normal distribution and may have a standard deviation indicating the degree of spread of the normal distribution. If CyRC' is obtained from Equation 2, and residuals are randomly generated according to a normal distribution with the same standard deviation as the residual distribution of the female fetus sample described above using Equation 1 and added to CyRC', an error occurs in chromosome Y of the male fetus sample. The allocated CyRC can be approximately obtained.

The step of generating virtual positive data from the virtual normal data may be determined as CXXYRC = CXXRC + CYRC when the target chromosome is an XXY chromosome. In the above formula, CXXYRC is the number of reads on target chromosome XXY in a chromosomal aneuploidy fetus, CXXRC is the number of reads on target chromosome XX in a normal female fetus, and CYRC is the number of reads assigned to chromosome Y.

The CYRC is a step of calculating the residual from CYRC = β ₀ + β ₁ TRC + β ₂ FF _snp + ε (Equation 3) and CYRC' = β ₀ + β ₁ TRC + β ₂ FF _snp (Equation 4) ; And it can be calculated from the step of calculating CYRC from the residual.

In Equation 3, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC, β ₂ is the coefficient between FF _snp and CYRC, and ε is the residual. In Equation 4 above, CYRC' is the predicted value of CYRC, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC', and β ₂ is the coefficient between FF _snp and CYRC'. It may include calculating CYRC from the calculated residual. Equations 3 and 4 can be modeled using male fetus (XY) samples. The distribution of the residuals can be assumed to follow a normal distribution and may have a standard deviation indicating the degree of spread of the normal distribution. Additionally, for samples with XXY chromosome aneuploidy, a regression model similar to that of male fetus samples and a normal distribution of residuals according to this model can be assumed. Since the number of reads (CYRC) assigned to chromosome Y is relatively small and negligible compared to the TRC, the TRC and FF of the female fetus (XX) sample can be regarded as the TRC and FF of the sample with XXY chromosome aneuploidy. By calculating CYRC' from Equation 4, using Equation 3 to randomly generate residuals according to a normal distribution with the same standard deviation as the residual distribution of male fetus samples described above, and adding them to CYRC', the CYRC of female fetuses can be approximated. You can get it.

The method includes establishing a chromosomal aneuploidy discrimination model from the produced hypothetical positive data and normal data.

The step of establishing a chromosomal aneuploidy discrimination model from the produced virtual positive data and normal data is, when the target chromosome is an autosome, CiRC (chromosome i read count) (i = 1 to 22), ATRC (Autosomal Total read count) ), FF, and GC content.

The step of establishing a chromosomal aneuploidy discrimination model from the produced virtual positive data and normal data is, when the target chromosome is a sex chromosome, CXRC (Chromosome , FF, GC content, or a combination thereof.

The chromosomal aneuploidy discrimination model can be trained by applying a machine learning algorithm. The machine learning algorithms include LGBM (Light Gradient Boosting Machine), AdaBoost, Voting Classifier, Random Forest, Logistic algorithm, Neural Network, and QDA. (Quadratic Discriminant Analysis). The LGBM is Guolin Ke et al., “LightGBM: A Highly Efficient Gradient Boosting Decision Tree”. Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS 2017), pp. It may be described in 3149-3157.

The method includes the step of detecting chromosomal aneuploidy in a test sample using a chromosomal aneuploidy discrimination model.

The step of detecting chromosomal aneuploidy of a test sample using the chromosomal aneuploidy discrimination model can be performed based on CiRC (i=1 to 22), ATRC, FF, and GC content when the target chromosome is an autosome. .

The step of detecting chromosomal aneuploidy of a test sample using the chromosomal aneuploidy determination model can be performed based on CXRC, CYRC, TRC, FF, GC content, or a combination thereof when the target chromosome is a sex chromosome.

Another aspect provides a computer-readable medium having a program applied to perform a method according to one aspect. The computer-readable media encompasses systems that include computer-readable media.

According to a method for detecting chromosomal aneuploidy in a fetus based on virtual data according to an embodiment, and a computer readable medium recording a program applied to perform the same, chromosomal aneuploidy in the fetus can be non-invasively diagnosed prenatally with excellent sensitivity and specificity. You can.

Figure 1a shows the distribution of virtual positive data and virtual negative data along three axes (ATRC, GC content, and C13RC) targeting chromosome 13 in three dimensions, and Figure 1b shows the distribution of virtual positive data and virtual negative data according to GC content and C13RC. The distribution of virtual positive data and virtual negative data is shown on a plane, and Figure 1c shows a model using virtual data, randomly mixed with data, trained using 80% of the data (training data), and 20% of the data. This is a result showing the sensitivity and positive predictive value of the test sample (test data) verified using , and Figure 1d shows the result of additional verification of the accuracy of the established model using a positive sample confirmed by amniocentesis.
Figure 2a shows the distribution of virtual positive data and virtual negative data along three axes (ATRC, GC, and C18RC) targeting chromosome 18 in three dimensions, and Figure 2b shows the distribution of virtual positive data and virtual negative data according to GC content and C18RC. The distribution of positive data and virtual negative data is shown on a plane, and Figure 2c shows a test in which data was randomly mixed in a model using virtual data, trained using 80% (learning data), and verified using 20%. This is a result showing the accuracy of the sample (test data), and Figure 2d shows the result of additional verification of the accuracy of the established model using a positive sample confirmed by amniocentesis.
Figure 3a shows the distribution of virtual positive data and virtual negative data along three axes (ATRC, GC content, and C21RC) targeting chromosome 21 in three dimensions, and Figure 3b shows the distribution of virtual positive data and virtual negative data according to GC content and C21RC. The distribution of virtual positive data and virtual negative data is shown on a plane, and Figure 3c shows a model using virtual data randomly mixed with data, trained using 80% (training data), and verified using 20%. This is a result showing the accuracy of one test sample (test data), and Figure 3d shows the result of additional verification of the accuracy of the established model using a positive sample confirmed by amniocentesis.
Figure 4a shows the hypothetical normal female fetus (XX) and the hypothetical benign data of monosomy X (XO) and triple The distribution is shown, and Figure 4b shows a hypothetical normal male fetus (XY) and hypothetical benign data, Klinefelter syndrome (XXY) and XYY syndrome (XYY) along two axes (GC and CYRC) targeting the male fetus ) shows the distribution of data, and Figure 4c is a test sample that was randomly mixed in a model using virtual positive data and virtual negative data, trained using 80% (learning data), and verified using 20%. This is a result showing the accuracy of (test data), and Figure 4d shows the result of additional verification of the accuracy of the established model using a positive sample confirmed by amniocentesis.

Hereinafter, the present invention will be described in more detail through examples. However, these examples are for illustrative purposes only and the scope of the present invention is not limited to these examples.

Example 1. Noninvasive detection of fetal chromosomal aneuploidy based on virtual data production

1. Preparation of samples

Blood from a total of 15,999 pregnant women was collected. Among these, amniotic fluid tests for karyotyping of the fetus were performed on those subjects who were detected as positive.

All subjects underwent standard prenatal aneuploidy screening at an accredited clinical trial site. First trimester screening includes serum pregnancy-associated plasma protein A (PAPP-A), total or free beta subunit of human chorionic gonadotropin (hCG), and fetal cervical zona pellucida. Includes measurement of nuchal translucency. Second trimester screening includes measurements of maternal serum alpha-fetoprotein (MSAFP), hCG, unconjugated estriol, and inhibin A.

As a result of karyotype analysis, 6 fetuses had aneuploidy on chromosome 13, 7 fetuses had aneuploidy on chromosome 18, 94 fetuses had aneuploidy on chromosome 21, 6 fetuses had an XO chromosome, and 9 fetuses had an XXY chromosome. Two fetuses had XYY chromosomes, and 13 fetuses had XXX chromosomes, and chromosomal aneuploidy was confirmed in the actual data.

2. Preparation of cell-free DNA and DNA libraries for DNA sequencing

Approximately 10 ml of peripheral blood was drawn from the subjects as described in 1. and collected in BCT™ tubes (Streck, Omaha, NE, USA). Collected blood samples were centrifuged at 1,200 x g for 15 minutes at 4°C. Blood plasma was collected and centrifuged again at 16,000 x g for 10 minutes at 4°C. Cell-free DNA (cfDNA) was obtained from centrifuged plasma using the MagListo™ cfDNA extraction kit (Bioneer, Korea).

The obtained cfDNA fragment was end-repaired using T4 DNA polymerase, Klenow DNA polymerase, and T4 polynucleotide kinase, and the cfDNA fragment was obtained again using Agencourt AMPure XP.

A DNA library for the ion proton sequencing system was prepared from the prepared cfDNA according to the protocol provided by the manufacturer (ThermoFisher Scientific). Using the 540 chip, an average sequencing coverage depth of 0.3x per nucleotide was calculated.

3. Massively parallel sequencing

DNA libraries prepared as described in 2 were massively parallel sequenced using the Ion Torrent S5XL™ system (ThermoFisher Scientific).

Different raw reads were obtained using Ion Torrent Suite™ software (ThermoFisher Scientific). The filtered reads were aligned to the human genome reference sequence hg19 by Burrows-Wheeler transform (BWT). Sequence reads that mapped to only one genomic location in hg19 were designated as unique reads. Of the total reads, approximately 3.3×10 ⁶ were unique, and the GC content of the total 15,999 samples ranged from approximately 39% to 45%.

Among the 15,999 samples described in 1., 348 male and 354 female fetus samples were selected with a unique read count of more than 2 million, a GC content of more than 39.5% and less than 42%, and a fetal DNA fraction of more than 4%. did. For the selected samples, the fetal DNA fraction was determined using SNP genotyping and imputation (Korean patent registration number KR 10-2031841). At this time, the MAF filtering standard value was 7%, and the correlation coefficient with the chromosome Y-based fetal DNA fraction was 98%.

To produce virtual normal (negative) data, male fetuses and female fetuses were separated from each other, two samples were randomly selected from each, combined, and randomly divided in half, so that no two samples overlapped. In other words, two samples were randomly selected from 348 male fetuses, combined, and then randomly divided in half to create virtual normal data for 5,000 new male fetuses. Here, when combining them into one and dividing them in half again, the number of leads corresponding to 50% of the total number of leads was randomly selected. The same method was performed on 354 samples from female fetuses to create virtual normal data of 5,000 female fetuses.

4. Production and detection of virtual positive data with chromosomal aneuploidy from virtual normal data

go. How to create virtual positive data with trisomy from virtual normal autosomal data

The method of creating a trisomy from an autosome is to add the result of multiplying the read count (RC) of autosome i by 50% of the fetal DNA ratio (FF). In other words, this can be expressed as TiRC = CiRC + CiRC × FF/2. In this formula, TiRC is the number of reads on target chromosome i in an aneuploidy fetus, CiRC is the number of reads on target chromosome i in a normal fetus, and FF is the fetal DNA fraction.

A specific method for producing chromosomal aneuploidy by targeting autosomes is explained using chromosomes 13, 18, and 21 as examples.

When a sample is randomly selected and chromosome 13 is targeted, the DNA reads of chromosome 13 are mixed with maternal and fetal reads. At this time, it is important to know the ratio of the fetus's DNA reads as accurately as possible. Therefore, for both male and female fetuses, the fetal DNA ratio (FF) using SNP-based prediction method (imputation) is used.

In the case of a normal fetus, the number of DNA reads assigned to chromosome 13 is determined by deriving from a pair of maternal chromosomes and a pair of fetal chromosomes. Therefore, the number of chromosomal DNA reads in a normal fetus is determined by multiplying the total number of DNA reads assigned to chromosome 13 by the fetal DNA ratio (FF). On the other hand, because trisomy 13 has one additional fetal chromosome, an additional 50% must be added to the normal number of fetal DNA reads. As a result, assuming that the number of DNA reads assigned to chromosome 13 of any sample is C13RC, chromosome 13 T13RC of the trisomy sample can be calculated as C13RC + C13RC × FF/2. Likewise, T18RC and T21RC can be obtained by applying this method to chromosomes 18 and 21.

me. Detection of positive data with trisomy on autosomes

(1) Detection of trisomy chromosome 13

When targeting chromosome 13, 10,000 virtual positive data with trisomy were produced from a total of 10,000 virtual normal data. Outliers that were statistically outside the normal range were excluded from the virtual data, and 9,823 virtual normal data and 9,646 virtual positive data were selected.

The results of the three-dimensional distribution of virtual positive data and virtual normal data along three axes (ATRC, GC content, and C13RC) targeting chromosome 13 are shown in Figure 1a. In addition, the distribution of virtual positive data and virtual normal data according to GC content and C13RC is shown in Figure 1b. Figure 1a shows the total read counts (ATRC) of DNA reads assigned to the autosome on chromosome 13 for trisomy 13 (T13) and normal chromosome 13, and The distribution between the total number of assigned DNA reads (C13RC) and GC content is shown. From these results, it was confirmed that ATRC, C13RC, and GC contents can serve as parameters for classifying T13 and normal samples. In particular, Figure 1b shows that C13RC can clearly distinguish between trisomy data and normal data according to changes in GC content within the limited range of ATRC (3.00 × 10 ⁶ to 3.05 × 10 ⁶ ).

9,823 virtual normal data and 9,646 virtual positive data are randomly mixed using three parameters of ATRC, C13RC, and GC content, and then trained using a logistic regression algorithm using 80% of the data (training data) to create a model. was established. The sensitivity and positive predictive value of the verified test sample (test data) were calculated using the remaining 20% of data, and the calculated results are shown in Figure 1c. As shown in Figure 1c, it was confirmed that the sensitivity and positive predictive value were each above 99.99% in the training and test data.

To confirm the accuracy of the established model, the results of additional verification using positive samples confirmed by amniocentesis are shown in Figure 1d. As shown in Figure 1d, when this model using virtual data was used, it was accurately predicted that the six actual positive samples (T13) confirmed through amniocentesis were positive data.

(2) Detection of trisomy 18 chromosome

In the case of chromosome 18, 10,000 trisomy virtual positive data were produced from 10,000 virtual normal data. Outliers were removed from the virtual data, and 9,823 virtual normal data and 9,529 false positive data were selected. Data were selected within the range of ATRC (3.00×10 ⁶ to 3.05×10 ⁶ ). Some data that were statistically outside the normal range were removed depending on the characteristics of each target chromosome.

The results of the three-dimensional distribution of virtual positive data and virtual normal data along three axes (ATRC, GC, and C18RC) targeting chromosome 18 are shown in Figure 2a. The results of the distribution of virtual positive data and virtual normal data according to GC content and C18RC are displayed on a plane, and are shown in Figure 2b. The model was established by randomly mixing the data using virtual data and then training it with a logistic regression algorithm using 80% of the data (training data). The results showing the sensitivity and positive predictive value of the test sample (test data) verified using the remaining 20% are shown in Figure 2c. To confirm the accuracy of the established model, the results of additional verification using positive samples confirmed by amniocentesis are shown in Figure 2d.

As shown in Figures 2a and 2b, ATRC, C18RC, and GC contents were able to distinguish between trisomy 18 (T18) and normal. Additionally, in Figure 2c, it was confirmed that the sensitivity and positive predictive value were 100% in the training data and test data, respectively. As shown in Figure 2d, when the method using virtual data was used, it was accurately predicted that the 7 actual positive samples (T18) confirmed through amniocentesis were positive data.

(3) Detection of trisomy 21 chromosome

In the case of chromosome 21, 10,000 trisomy virtual positive data were produced from 10,000 normal virtual data. Outliers were removed from the virtual data, and 9,823 virtual normal data and 9,594 virtual positive data were selected. Data were selected within the range of ATRC (3.00×10 ⁶ to 3.05×10 ⁶ ). Some data that were statistically outside the normal range were removed depending on the characteristics of each target chromosome.

The results of the three-dimensional distribution of virtual positive data and virtual normal data along three axes (ATRC, GC, and C21RC) targeting chromosome 21 are shown in Figure 3a. The results of displaying the distribution of virtual positive data and virtual normal data according to GC content and C21RC on a plane are shown in Figure 3b. The model was established by randomly mixing the data using virtual data and then training it with a logistic regression algorithm using 80% of the data (learning data). The results showing the sensitivity and positive predictive value of the test sample (test data) verified using the remaining 20% are shown in Figure 3c. To confirm the accuracy of the established model, the results of additional verification using positive samples confirmed by amniocentesis are shown in Figure 3d.

As shown in Figures 3a and 3b, the ATRC, C18RC, and GC contents were able to distinguish between trisomy 21 (T21) and normal. Additionally, in Figure 3c, it was confirmed that the sensitivity and positive predictive value were over 99.8% in the training and test data. As shown in Figure 3d, when the method using virtual data was used, it was accurately predicted that 94 actual positive samples (T21) confirmed through amniocentesis were positive data.

all. How to create virtual positive data with sex chromosome aneuploidy from virtual normal sex chromosome data

To create data with fetal sex chromosome aneuploidy, male fetus (XY) data, female fetus (XX) data, and fetal DNA ratio (FF) were used.

For FF, chromosome Y-based FF was used for male fetus samples, and machine learning-based FF was used for female fetus samples.

(1) Detection of XO sex chromosome aneuploidy

The method for creating virtual positive data representing It is done. However, there is a certain amount of misassigned reads on chromosome Y.

CXRC = CXYRC - CYRC + CyRC

In the above formula, CXRC is the number of reads on target chromosome XO in a chromosomal aneuploidy fetus, CXYRC is the number of reads on target chromosome This is the number of incorrectly assigned leads.

The number of reads incorrectly assigned to chromosome Y (CyRC) was expressed in a regression equation proportional to the total number of reads (total read size).

That is, it is expressed as the regression equation in Equation 1.

CyRC = β ₀ + β ₁ TRC + ε (Equation 1)

In Equation 1, TRC is the number of reads of the entire chromosome, β ₀ is the intercept, β ₁ is the coefficient between TRC and CyRC, and ε is the residual.

Equation 1 uses TRC as the independent variable. The predicted value of the dependent variable can be expressed in Equation 2 below.

CyRC' = β ₀ + β ₁ TRC (Equation 2),

In Equation 2, CyRC' is the predicted value of CyRC, TRC is the number of reads of the entire chromosome, β ₀ is the intercept, and β ₁ is the coefficient between TRC and CyRC'.

The formula is calculated using 8,737 female fetal samples as follows.

CyRC' = 3.8×10 ^-5 TRC + 8.7

The residual was calculated from Equation 1 and Equation 2 above. The distribution of residuals can be assumed to follow a normal distribution and can have a standard deviation that indicates the spread of the normal distribution. The standard deviation of the residuals of 14 was calculated from the normal distribution. A new residual was calculated using a normal distribution according to this standard deviation. Calculate CyRC' from the total read count (TRC) of random male fetuses and Equation 2, and calculate the new CyRC by adding the new residual calculated previously. As a result, CXRC can be obtained by removing the read counts (chromosome Y read counts: CYRC) assigned to chromosome Y from a random male fetus sample CXYRC and adding the newly obtained CyRC.

(2) Detection of XXX sex chromosome aneuploidy

The way to determine the number of reads for a fetus with XXX sex chromosome aneuploidy is to add the number of reads (CXXRC) multiplied by 50% of FF from the number of reads in a random female fetus sample (CXXRC). In other words, it can be calculated as CXXXRC = CXXRC + CXXRC × FF/2. In the above formula, CXXXRC is the number of reads on target chromosome XXX in a chromosomal aneuploidy fetus, CXXRC is the number of reads on target chromosome XX in a normal female fetus, and FF is the fetal DNA fraction.

(3) Detection of XYY sex chromosome aneuploidy

The method for calculating the read count of a fetus with XYY sex chromosome aneuploidy is to calculate the read count (chromosome ) is added. In other words, the read counts (chromosome XYY read counts: CXYYRC) of fetuses with That is, it can be calculated using the following formula.

CXYYRC = CXYRC + CYRC - CyRC

In the above formula, CXYYRC is the number of reads on target chromosome XYY in a chromosomal aneuploidy fetus, CXYRC is the number of reads on target chromosome This is the number of reads incorrectly assigned to target chromosome Y.

Here, CyRC was calculated using equations 1 and 2 and the standard deviation of the residuals of 14.

(4) Detection of XXY sex chromosome aneuploidy

The chromosome XXY read count (CXXYRC) of a fetus with XXY sex chromosome aneuploidy can be calculated as CXXYRC = CXXRC + CYRC. In the above formula, CXYYRC is the number of reads on target chromosome XYY in a chromosomal aneuploidy fetus, CXYRC is the number of reads on target chromosome This is the number of reads incorrectly assigned to target chromosome Y.

First, CYRC was expressed as the regression equation in Equation 3 using chromosome Y-based FF.

CYRC = β ₀ + β ₁ TRC + β ₂ FF _snp + ε (Equation 3)

In Equation 3, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC, β ₂ is the coefficient between FF _snp and CYRC, and ε is the residual.

Equation 3 used TRC and chromosome Y-based FF as independent variables. The predicted value of the dependent variable can be expressed in Equation 4 below.

CYRC' = β ₀ + β ₁ TRC + β ₂ FF _snp (Equation 4),

In Equation 4 above, CYRC' is the predicted value of CYRC, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC', and β ₂ is the coefficient between FF _snp and CYRC'.

The formula is calculated using 9,406 male fetal samples as follows.

CYRC' = 0.00018 TRC + 72.5 FF _snp - 576

From Equations 3 and 4 above, 92 was calculated as the standard deviation of the residual, and a new residual was calculated by normal distribution according to this standard deviation. Calculate CYRC' from Equation 4 using the total read count (TRC) of random female fetuses and SNP-based FF, and calculate the new CYRC by adding the new residuals obtained previously. As a result, CXXYRC can be obtained by adding the newly obtained CYRC to a random female fetus sample CXXRC.

la. Detection of positive data with sex chromosome aneuploidy

Virtual normal female fetus (XX) along two axes (GC and CXRC) targeting XX chromosomes of female fetus within the range of total read counts (TRC) (3.00×10 ⁶ to 3.05×10 ⁶ ). And a graph representing the distribution of hypothetical positive data, monosomy X (XO), and triple X syndrome (XXX) samples, is shown in Figure 4a. As shown in Figure 4a, in the case of the targeted female fetal X chromosome (XX), CXXRC, CXORC, and CXXXRC, the number of reads assigned to chromosome

Klein, a virtual normal male fetus (XY) and virtual benign data along two axes (GC content and CYRC) targeting the XY chromosomes of male fetuses within the range of TRC (3.00×10 ⁶ to 3.05×10 ⁶ ). A graph representing the distribution of Felter syndrome (XXY) and XYY syndrome (XYY) data on a plane is shown in Figure 4b. As shown in Figure 4b, in the case of the target male fetal chromosome Y (XY), CXYRC, CXXYRC, and CXYYRC, the number of reads assigned to chromosome Y (CYRC) was distributed differently depending on the GC content.

Due to these differences in distribution, it was confirmed that CXRC, CYRC, and GC contents can be used as parameters to classify normal fetal samples such as XX and XY and sex chromosome aneuploidy samples with XO, XXX, XXY, or XYY.

Using 10,000 virtual normal male fetuses and 9,994 virtual normal female fetus samples, 10,000 XYY, 10,000 XO, 9,957 XXX, and 9,967 XXY virtual positive data were created, respectively. In addition, an additional 10,000 virtual normal male fetuses and 9,930 virtual normal female fetuses were created and used so as not to overlap with those mentioned above. Here, the virtual data was randomly mixed and trained using 80% to establish a model, and the accuracy of the test sample verified using the remaining 20% is shown in Figure 4c. As shown in Figure 4c, it was confirmed that the training and test data showed an accuracy of over 99.8% in sensitivity and positive predictive value, respectively.

The results of further verification of the accuracy of the established model using positive samples confirmed by amniocentesis are shown in Figure 4d. As shown in Figure 4d, the actual chromosomal aneuploidy (positive) samples, which were 6 XO, 9 XXY, 2

Claims (19)

As a method for detecting chromosomal aneuploidy in a fetus,
Obtaining sequence information (Read) of a plurality of nucleic acid fragments obtained from a biological sample of a woman pregnant with a normal fetus;
Mapping the obtained sequence information to a human reference genome and assigning the sequence information of the nucleic acid fragment to a chromosome;
Based on the sequence information of the nucleic acid fragment specified in the chromosome, read count (RC), fetal DNA fraction (FF), and GC content of the sequence information of the nucleic acid fragment in the target chromosome are calculated in a normal fetus. steps;
A step of producing virtual normal data from normal data based on the number of reads in the target chromosome and fetal DNA fraction in the normal fetus,
The normal data is the number of reads in the target chromosome of a normal fetus;
Generating virtual positive data from the virtual normal data,
The virtual positive data is the number of reads in the target chromosome of a fetus with chromosomal aneuploidy;
Establishing a chromosomal aneuploidy discrimination model from the produced virtual normal data and virtual positive data; and
A method comprising the step of detecting chromosomal aneuploidy of a test sample using the chromosomal aneuploidy determination model. The method of claim 1, wherein the target chromosome is selected from the group consisting of chromosomes 1 to 22, chromosome X, and chromosome Y. The method of claim 2, wherein the target chromosome is selected from the group consisting of chromosome 13, chromosome 18, and chromosome 21. The method of claim 1, wherein the step of producing virtual normal data from the normal data includes
A method of randomly selecting two or more normal samples, combining the selected fetal samples, and randomly dividing them again to generate virtual normal data. The method of claim 1, wherein the step of producing virtual normal data from the normal data includes
Distinguishing between normal male fetus samples and normal female fetus samples; and
Randomly select two or more samples from normal male fetus samples, combine the selected male fetus samples, and randomly divide again to generate virtual normal male fetus data, or
A method comprising the step of randomly selecting two or more samples from normal female fetus samples, combining the selected female fetus samples, and randomly dividing them again to generate virtual normal female fetus data. The method of claim 1, wherein the step of generating virtual positive data from the virtual normal data is determined by the following equation when the target chromosome is an autosome:
TiRC = CiRC + CiRC × FF/2,
In the above equation,
TiRC is the number of reads on target chromosome i in aneuploidy fetuses,
CiRC is the number of reads on target chromosome i in a normal fetus;
FF is the fetal DNA fraction. The method of claim 1, wherein the step of generating virtual positive data from the virtual normal data is determined by the following equation when the target chromosome is the XXX chromosome:
CXXXRC = CXXRC + CXXRC×FF/2,
In the above equation,
CXXXRC is the number of reads on target chromosome XXX in aneuploidy fetuses,
CXXRC is the number of reads on target chromosome XX in a normal female fetus,
FF is the fetal DNA fraction. The method of claim 1, wherein the step of generating virtual positive data from the virtual normal data is determined by the following equation when the target chromosome is an XO chromosome:
CXRC = CXYRC - CYRC + CyRC,
In the above equation,
CXRC is the number of reads on target chromosome XO in aneuploidy fetus,
CXYRC is the number of reads on target chromosome XY in a normal male fetus,
CYRC is the number of reads on target chromosome Y,
CyRC is the number of reads incorrectly assigned to target chromosome Y. The method of claim 1, wherein the step of generating virtual positive data from the virtual normal data is determined by the following equation when the target chromosome is an XYY chromosome:
CXYYRC = CXYRC + CYRC - CyRC,
In the above equation,
CXYYRC is the number of reads on target chromosome XYY in aneuploidy fetus,
CXYRC is the number of reads on target chromosome XY in a normal male fetus,
CYRC is the number of reads assigned to chromosome Y in a normal male fetus,
CyRC is the number of reads incorrectly assigned to target chromosome Y. The method of claims 8 and 9, wherein the CyRC
Steps for calculating residuals from Equations 1 and 2 below:
CyRC = β ₀ + β ₁ TRC + ε (Equation 1),
In Equation 1, TRC is the number of reads of the entire chromosome, β ₀ is the intercept, β ₁ is the coefficient between TRC and CyRC, and ε is the residual,
CyRC' = β ₀ + β ₁ TRC (Equation 2),
In Equation 2, CyRC' is the predicted value of CyRC, TRC is the number of reads of the entire chromosome, β ₀ is the segment, and β ₁ is the coefficient between TRC and CyRC'; and
A method calculated from the step of calculating CyRC from the residual. The method of claim 1, wherein the step of generating virtual positive data from the virtual normal data is determined by the following equation when the target chromosome is an XXY chromosome:
CXXYRC = CXXRC + CYRC,
In the above equation,
CXXYRC is the number of reads on target chromosome XXY in aneuploidy fetus,
CXXRC is the number of reads on target chromosome XX in a normal female fetus,
CYRC is the number of reads assigned to chromosome Y. The method of claim 11, wherein the CYRC calculates a residual from the following equations 3 and 4:
CYRC = β ₀ + β ₁ TRC + β ₂ FF _snp + ε (Equation 3),
In Equation 3, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC, β ₂ is the coefficient between FF _snp and CYRC, and ε is the residual,
CYRC' = β ₀ + β ₁ TRC + β ₂ FF _snp (Equation 4),
In Equation 4 above, CYRC' is the predicted value of CYRC, TRC is the number of reads of the entire chromosome, FF _snp is SNP-based FF, β ₀ is the intercept, β ₁ is the coefficient between TRC and CYRC', and β ₂ is the coefficient between FF _snp and CYRC'; and
A method calculated from the step of calculating CYRC from the residual. The method of claim 1, wherein the step of establishing a chromosomal aneuploidy discrimination model from the produced virtual normal data and virtual positive data includes
When the target chromosome is an autosome, the method is performed based on CiRC (chromosome i read count) (i = 1 to 22), ATRC (Autosomal Total read count), FF, GC content, or a combination thereof. The method of claim 1, wherein the step of detecting chromosomal aneuploidy of a test sample using a chromosomal aneuploidy determination model is performed by determining CiRC (i=1 to 22), ATRC, FF, and GC content, or these, when the target chromosome is an autosome. A method that is performed based on a combination of. The method of claim 1, wherein the step of establishing a chromosomal aneuploidy discrimination model from the produced virtual normal data and virtual positive data includes
When the target chromosome is a sex chromosome, the method is performed based on Chromosome X read count (CXRC), Chromosome Y read count (CYRC), Total read count (TRC), FF, GC content, or a combination thereof. The method of claim 1, wherein the step of detecting chromosomal aneuploidy of a test sample using a chromosomal aneuploidy determination model is performed based on CXRC, CYRC, TRC, FF, and GC contents, or a combination thereof, when the target chromosome is a sex chromosome. How to do it. The method according to claim 1, wherein the chromosomal aneuploidy discrimination model is trained by applying a machine learning algorithm. The method of claim 17, wherein the machine learning algorithm is LGBM (Light Gradient Boosting Machine), AdaBoost, Voting Classifier, Random Forest, Logistic algorithm, and Artificial Neural Network. Network), and QDA (Quadratic Discriminant Analysis). A computer-readable medium recording a program applied to perform the method according to any one of claims 1 to 18.