AU2018428853A1 - Methods and compositions for the analysis of cancer biomarkers - Google Patents

Abstract

Described herein are improved methods, compositions, and kits for analysis of minimal residual solid tumor.

Description

METHODS AND COMPOSITIONS FOR THE ANALYSIS OF CANCER

BIOMARKERS

BACKGROUND

[0001] Molecular tests can detect residual disease after a treatment. The presence of residual disease indicates that the treatment did not completely eliminate a tumor, where treatment may include surgery, radiotherapy, chemotherapy, endocrine therapy, or targeted molecular therapy.

[0002] Following surgical treatments, positive surgical margins are defined as tumor cells on the surface of an excised tissue specimen. Since the surface of the excised specimen is topologically equivalent to the wall of the incision, tumor cells on the surface of the incision indicate the presence of residual tumor in a patient after surgical treatment.

[0003] Following medical treatments, Pathologic Complete Response (pCR) is defined as the absence of residual tumor in tissue from patients who were previously diagnosed with invasive cancer. pCR is used as a primary endpoint to determine the success of emerging breast cancer treatments in the neoadjuvant setting. Innovative clinical trial designs have validated pathologic complete response (pCR) as a surrogate endpoint, and are now validating pCR as a therapeutic endpoint.

SUMMARY

[0004] Described herein are methods and compositions that are useful for an improved RNA- based test suitable for analysis of tumor margins from surgical samples for residual disease, or for analysis of residual disease in post-treatment cancer patients from other samples.

[0005] In some aspects, the disclosure provides a method of distinguishing a cancer from adjacent healthy tissue, said method comprising: (a) obtaining a specimen from a human subject, (b) detecting a presence of a set of markers in said specimen by performing an amplification reaction in a plurality of polynucleotides from said specimen, wherein said set of markers is selected from the group consisting essentially of: Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain

( COL10A1 ); and (c) distinguishing said cancer when a threshold level of said set of markers is detected. In some instances, said plurality of polynucleotides comprise RNA, cDNA, or DNA. In some instances, the detecting comprises using a DNA-intercalating dye or a fluorescent probe, such as a TaqMan probe. In some instances said amplification reaction is a PCR reaction, such as a qPCR reaction or an RTqPCR reaction. In some instances, said method can distinguish said cancer in at least lOng of said plurality of polynucleotides from specimen. In some instances, said method can distinguish said cancer in at least 250 cells of said specimen. In some instances, said amplification reaction uses at least one primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356, for example to convert RNA into cDNA and/or to amplify a cDNA. In some instances, the specimen is a frozen specimen, a fresh specimen, or a fixed specimen. In some instances, the specimen is a biopsy specimen, such as a liquid biopsy, a solid tissue biopsy, or a surgical excision. In some instances, said specimen is obtained by imprint cytology, with for example a touch- preparation. In some instances said specimen is obtained by scrape preparation, a nipple aspiration, or a ductal lavage. In some instances said cancer is breast cancer, including, but not-limited to, invasive adenocarcinoma, invasive ductal breast cancer, and invasive lobular breast cancer. In some instances, said method distinguishes said breast cancer from adjacent healthy tissue with greater than 90% accuracy, greater than 90% sensitivity, or greater than 90% specificity. In some instances, said method quantitates an amount of said cancer. In some instances said method further comprises outputting a percentage of said plurality of polynucleotides expressing said markers from said specimen. In some instances, the method further comprises comparing said set of markers from said specimen to said set of markers from said control specimen, such as a second specimen from said human subject or a synthetic nucleotide control. In some instances, the method further comprises performing a second assay to distinguish said cancer, such as an immunohistochemistry assay. In some instances, said threshold level of said MMP I / is 1,000 copies per microliter, said threshold level of said IBSP is 25 copies per microliter, and said threshold level of said COL10A1 is 700 copies per microliter. In some instances, said set of markers is selected from the group consisting of: Matrix Metallopeptidase 11 (MMP 11), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain (COLIOAI). In some aspects, said amplification reaction can be a singleplex reaction or a multiplex reaction.

[0006] In some aspects, the disclosure provides a kit comprising, at least one primer sequence that has at least 90% identity to any one of SEQ ID NO: 1- SEQ ID NO: 356, and a buffer system. In some instances said buffer system is a PCR buffer system. In some instances, the kits further comprise a DNA-intercalating dye, a fluorescent probe, such as a TaqMan compatible probe. In some instances the kit also comprises a negative control sample, a positive control sample, or a synthetic nucleotide control. [0007] In some aspects, the disclosure provides isolated nucleic acid comprising a primer sequence that has at least 90%, at least 95%, or at least 99% identity to SEQ ID NO: 1- SEQ ID NO: 356.

[0008] In some aspects the disclosure provides a method of identifying a biomarker for a cancer comprising: (a) analyzing, by a computer system, a cohort of biomarkers from a population of subjects afflicted with a cancer; (b) applying, by said computer system, a first filter to said cohort of said biomarkers to identify a first subset of biomarkers from said cohort that has at least a 3 -fold higher expression level in said cancer as compared to a healthy control biomarker; (c) applying, by said computer system, a second filter to said first subset of biomarkers to identify a second subset of biomarkers that have a false discovery rate for said cancer that is less than 0.000001; and (d) applying, by said computer system, a correlation based filter selection to said second subset of biomarkers to identify the biomarkers that classify the largest number of different types of said cancer. In some aspects, said correlation based filter is an anti-correlation based method. In some aspects, the method further comprises using the identified biomarkers as features input into a machine learning algorithm that distinguishes clinical specimens based on predefined attributes. In some aspects said cancer is breast cancer, including, but not-limited to invasive adenocarcinoma, invasive ductal breast cancer, and invasive lobular breast cancer. In some aspects, said one or more biomarkers identify said cancer with greater than 90% accuracy, greater than 90% sensitivity, or greater than 90% specificity. In some aspects, said one or more biomarkers are therapeutic targets. In some aspects, said false discovery rate is a /i-value for said cancer that is less than 0.0000001.

[0009] Additional aspects and advantages of the present disclosure will become readily apparent to those skilled in this art from the following detailed description, wherein only illustrative embodiments of the present disclosure are shown and described. As will be realized, the present disclosure is capable of other and different embodiments, and its several details are capable of modifications in various obvious respects, all without departing from the disclosure. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.

INCORPORATION BY REFERENCE

[0010] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference.

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments, in which the principles of the invention are utilized, and the accompanying drawings (also“Figure” and“FIG.” herein), of which:

[0012] FIGURE 1 is a diagram illustrating positive versus clear surgical margins.

[0013] FIGURE 2A is a Volcano plot of 20,253 mRNAs in 1,014 samples. RNA Seq was used to analyze 1,014 samples from early-stage tumors and healthy samples from adjacent tissue. Selected genes had the highest Correlation-based Feature Selection scores among genes that passed /i- value threshold (dashed horizontal line) and fold-change threshold (dashed vertical line).

[0014] FIGURE 2B panels (a-c) are cumulative frequency plots of 1,536 patient samples that show that a 3-gene set ( MMP11 , COL10AJ IBSP ) is overexpressed in samples from early-stage tumors and adjacent healthy tissue. The genes have comparable distributions on RNA Seq samples (a), a subset of samples that were also analyzed by microarray (b), and a subset that were also analyzed by RTqPCR (c). These results confirm that expression is not platform-specific. Panels (c-e) are 2D-Density maps illustrating the advantage of a multianalyte test over a single biomarker. Separation of tumor and healthy improves as we progress from RNA Seq to Microarray to custom RTqPCR.

[0015] FIGURE 3 is a chart showing a Principal Component Analysis (PCA) of all available microarray probes shows a clear demarcation between tumor (left dots) and healthy samples (right dots).

[0016] FIGURE 4 depicts receiver-operator characteristic (ROC) curves of classifiers for a 3-gene set including MMPH, COL10AJ IBSP. ROC curves show the tradeoff between sensitivity and specificity over all possible thresholds. The solid dark line shows performance of the 3-gene test on 939 cross-validated RNA Seq samples.

[0017] FIGURE 5 illustrates an error plot of a 3-gene set ( MMP11 , COL10AJ IBSP ) in 939 RNA Seq samples. In contrast to ROC plots, which show the tradeoff between sensitivity and specificity, error plots set the threshold based on the tradeoff between Type I and Type II errors. Type I errors (False Positives) trigger unnecessary re-excisions. Type II errors (False Negatives) indicate positive margins that were not detected. We use these tradeoffs to guide our threshold selection.

[0018] FIGURE 6 is a graph illustrating that negative controls to detect overfitting demonstrate that predictive models were correctly cross-validated (n = 939 RNA Seq samples).

[0019] FIGURE 7A and FIGURE 7B depict charts showing analytic validation of qPCR assays for using clinical-grade reagents. FIGURE 7A panel a depicts amplification plots of 20 microliter qPCR reactions. 12 concentrations of synthetic cDNA template 1.1 million copies per microliter to 0 copies per microliter), including lO-fold dilutions for 6 high concentrations (5 technical replicates) and 2-fold dilutions for 5 low concentrations (7 technical replicates). One concentration point overlapped in the high and low concentration series. Each primer pair includes 24 replicates of no-template controls. Error bars at each cycle represent 95% Cl of technical replicates.

[0020] FIGURE 7A panel b depict fluorescence versus cycle plots to determine Ct for MMP11. A 4-parameter linear model was fitted to 5 technical replicates (circles). The maximum of the second derivative was used to define the Ct (CtD2).

[0021] FIGURE 7B panel c depicts threshold cycle versus template dilution plots to calculate linear range. The linear range is defined as the range of concentrations where CtD2 fit a straight line with R-squared >0.995. Red lines indicate 95% Confidence Intervals calculated from 200 bootstraps. FIGURE 7B panel d depicts melt plots confirm to specificity of the primers. Increasing temperature denatures PCR amplicons, which decreases

fluorescence. A single peak of the negative first derivative confirms the presence of a single amplicon. The peak corresponds to the expected melting temperature (dashed line).

[0022] FIGURE 7A and FIGURE 7B panels e-h depict charts showing analytic validation of qPCR assays for IBSP RNA as for MM PI / All assays used clinical-grade reagents. Panel e depicts amplification plots of 20 microliter qPCR reactions. 12 concentrations of synthetic cDNA template (1.1M to 0 copies per microliter), including lO-fold dilutions for 6 high concentrations (5 technical replicates) and 2-fold dilutions for 5 low concentrations (7 technical replicates). One concentration point overlapped in the high and low concentration series. Each primer pair includes 24 replicates of no-template controls. Error bars at each cycle represent 95% Confidence Intervals of technical replicates.

[0023] FIGURE 7A Panel f depicts fluorescence versus cycle plots to determine Ct for IBSP. A 4-parameter linear model was fitted to 5 technical replicates (circles). The maximum of the second derivative was used to define the Ct (CtD2). FIGURE 7B panel g depicts threshold cycle versus template dilution plots to calculate linear range. The linear range is defined as the range of concentrations where CtD2 fit a straight line with R-squared >0.995. Red lines indicate 95% Confidence Intervals calculated from 200 bootstraps.

[0024] FIGURE 7B panel h depict melt plots that demonstrate the specificity of the primers. Increasing temperature denatures PCR amplicons, which decreases fluorescence. A single peak of the negative first derivative confirms the presence of a single amplicon. The peak corresponds to the expected melting temperature (dashed line).

[0025] FIGURE 7A and FIGURE 7B panels i-1 depict analytic validation of qPCR assays for COL10A1 RNA as for MMP I / All assays use clinical-grade reagents. FIGURE 7A panel i depict amplification plots of 20 microliter qPCR reactions. 12 concentrations of synthetic cDNA template (1.1M to 0 copies per microliter), including lO-fold dilutions for 6 high concentrations (5 technical replicates) and 2-fold dilutions for 5 low concentrations (7 technical replicates). One concentration point overlapped in the high and low concentration series. Each primer pair included 24 replicates of no-template controls. Error bars at each cycle represent 95% Confidence Intervals of technical replicates. FIGURE 7A Panel j depicts fluorescence versus cycle plots to determine Ct for COL10A1. A 4-parameter linear model was used to fit all 5 technical replicates (circles). The maximum of the second derivative (green curve) was used to define the Ct (CtD2). FIGURE 7 A panel k depicts threshold cycle versus template dilution plots to calculate linear range. The linear range is defined as the range of concentrations where CtD2 fit a straight line with R-squared >0.995. Red lines indicate 95% Confidence Intervals calculated from 200 bootstraps. Panel 1 depicts melt plots confirm to specificity of the primers. Increasing temperature denatures PCR amplicons, which decreases fluorescence (black line). A single peak of the negative first derivative (red line) confirms the presence of a single amplicon. The peak corresponds to the expected melting temperature (dashed line).

[0026] FIGURE 8A, FIGURE 8B, and FIGURE 8C are graphs depicting absolute quantification (RT-qPCR) of the 3 RNAs in the 3-gene set ( MMP11 , COL10AJ IBSP ) in 22 patient samples using Tukey Boxplots. Tukey Boxplots: the thick center line represents the mean, boxes show the interquartile range (Q1-Q3). Cumulative Frequency plots show the distribution of expression in tumor and healthy samples. Panel b depicts absolute

quantification (RTqPCR) of RNAs in 22 patient samples using density plots. Density plots illustrate the advantage of combining multiple biomarkers. Copy numbers are adjusted for tumor percent because each tumor specimen contains a differing amount of healthy cells. [0027] FIGURE 9 is a graph depicting a Receiver-Operator Characteristic (ROC) Curve of the 3-Gene Classifier. ROC curves show the tradeoff between sensitivity and specificity over all possible thresholds. The 3-gene classifier uses Random Forest to distinguish between tumor and adjacent healthy tissue. Performance estimates are based on 5-fold cross validation of 22 samples that were analyzed with the disclosed RTqPCR assays.

[0028] FIGURE 10 depicts a plot showing Generalized Linear Model (glm) (dashed line) sample discrimination using IBSP RNA in 22 patient samples, analyzed by the disclosed RTqPCR assay. The disclosed RTqPCR assays can resolve a greater difference in analytes than RNA Seq. The disclosed assays perform so well that a simple linear model can correctly classify 100% of the analyzed samples using a single biomarker. In contrast, RNA Seq required a complex combination of 3 biomarkers, and still did not achieve 100% accuracy.

[0029] FIGURE 11 depicts a plot showing Generalized Linear Model (glm) (dashed line) sample discrimination using MMP11 RNA in 22 patient samples, analyzed by the disclosed RTqPCR assay.

[0030] FIGURE 12 shows a chart depicting a Tumor Probability Score calculated using the 3 -gene classifier described in EXAMPLE 1. The 3 -gene classifier uses the Random Forest algorithm to calculate a Tumor Probability Score (T) from zero to one. Panel a shows the T score for RNA Seq samples from 901 tumors (black) and 113 adjacent healthy samples (grey). Panel b shows the T score for RTqPCR samples from 11 tumors (black) and 11 adjacent healthy samples (grey).

[0031] FIGURE 13 shows a computer system that is programmed or otherwise configured to implement methods provided herein.

DETAILED DESCRIPTION

I. Overview of Pathologic Complete Response

[0032] While various embodiments of the invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed.

[0033] pCR has quickly become the primary endpoint for -50% of enrolling phase II rectal cancer trials, and 45% of phase III preoperative breast cancer trials. Unpublished results from the I-SPY 2 TRIAL of high-risk breast cancer patients indicate that pCR was statistically associated with 3-year outcomes on pooled patients across all treatment arms. After 3 years, patients who achieved pCR had a 6% recurrence risk (event-free survival), compared to 24% recurrence risk for those who did not achieve pCR.

[0034] Improving surrogate endpoints will help to replace treatment regimens with ones that are more effective, less toxic, and that improve survival. However, existing technologies are subjective, qualitative, and underpowered because they are based on visual analysis of a limited number of tissue sections. Moreover, pCR is labor intensive and currently only provided by specialty clinical centers as part of research protocols. Pathology labs routinely examine 3-5 microscopic tissue sections. If therapeutic response is ultimately verified as a therapeutic goal, busy pathology practices will be overwhelmed by requests to examine thousands of sections from hundreds of thousands of U.S. patients with invasive breast cancer. Described herein is a quantitative molecular analysis of residual tumor for identifying improved treatment regimens and complete excision of malignant tissue from patients.

II. Overview of Positive Surgical Margins

[0035] Most U.S. breast cancer patients are treated with breast conservation surgery

(lumpectomy), where the goal is to remove the entire tumor, bounded by a thin margin of healthy tissue (FIG. la). Positive margins are defined as malignant cells that touch the cut surface of a specimen (FIG.lb), indicating residual tumor in the bed of the incision. Positive margins increase the risk of recurrence and disease-specific mortality. As an example, in a cohort of 1,043 consecutive patients, positive margins were the strongest risk factor of disease-specific mortality among patients with early-stage breast tumors: the lO-year risk of death from breast cancer was 3.9x higher for patients with positive margins, relative to patients with negative margins (95% Cl: 1.4-11.5, p = 0.011). See , e.g ., Meric F, Mirza NQ, Vlastos G, Buchholz TA, Kuerer HM, Babiera GV, Singletary SE, Ross MI, Ames FC, Feig BW, Krishnamurthy S, Perkins GH, McNeese MD, Strom EA, Valero V, Hunt KK. Positive surgical margins and ipsilateral breast tumor recurrence predict disease-specific survival after breast-conserving therapy, Cancer, 2003 Feb l5;97(4):926-33.

[0036] Patients with positive margins have a higher risk of recurrence (HR: 2.52, 95%CI: 1.04-6.09) than patients with 10 positive lymph nodes (HR: 2.32, 95% Cl: 1.29-4.14). These findings hold, even under modern treatment protocols that include localized radiation, endocrine therapy, targeted molecular therapy, and the option of systemic chemotherapy. Detecting and treating positive margins is important because the risk of recurrence typically cannot be mitigated by additional chemotherapy or a radiation boost. Obtaining clear margins is a canon of surgical oncology, and is codified in clinical guidelines (ASCO and NCCN) and consensus statements (SSO and ASRO).

[0037] There is a need to improve the evaluation of surgical margins. Histopathology has been the best way to examine tumors for over a century, but it is not an ideal way to hunt for residual disease on the surface of a specimen. A retrospective analysis of 1,201 lumpectomy margins from Harvard’s Brigham and Women’s Hospital found that when microscopy was used to detect positive margins, it had a 51% sensitivity, 69.5% specificity, 19% false negative rate, and 65% false positive rate. See , e.g ., Tang R, Coopey SB, Specht MC, Lei L, Gadd MA, Hughes KS, Brachtel EF, Smith BL. Lumpectomy specimen margins are not reliable in predicting residual disease in breast conserving surgery. Am J Surg. 2015

Jul;2l0(l):93-8. These results were consistent with a prospective, randomized-control trial at Yale, where microscopy of the primary specimen had a false negative rate of 20%.

Undersampling is likely to be a primary culprit; microscopic sections only sample a small portion of a specimen’s surface. Some pathologists therefore conclude that margin analysis is the weak link in breast cancer care.

[0038] Many have tried to reduce reexcisions by testing margins during an operation, but these technologies have failed to reach a clinical impact. This is primarily due to the preliminary nature of rapid intraoperative test results— surgeons use them to predict post- operative test results. Since the relevant reference-standard has a 51% sensitivity and 70% specificity, test discordance has created an insurmountable barrier for adoption— even a perfect intraoperative test cannot predict which margins pathology will call positive.

Accordingly, we describe herein a method using nucleic acid tests to improve post-operative testing.

[0039] Improved testing has the potential to reduce Type I & II Errors. Type I errors are known as false positives. False positives have proven a significant barrier in the adoption of analysis of tumor margins by microscopy /histology; in a previous study of lumpectomy margin analysis by Tang et al. only 149 (32%) of 462 positive microscopy results actually had residual tumor along the margin. See , e.g., Tang R, Coopey SB, Specht MC, Lei L, Gadd MA, Hughes KS, Brachtel EF, Smith BL. Lumpectomy specimen margins are not reliable in predicting residual disease in breast conserving surgery. Am J Surg. 2015 Jul;2l0(l):93-8.

[0040] The 313 false positives triggered an alarming number of unnecessary surgeries. Type II errors are known as false negatives; in the same Tang et al. study, false negatives also presented a problem as traditional microscopy only detected 149 (51%) of the 293 margins that contained residual disease. The 144 patients with false negative results had a high risk of recurrence and mortality, which could have been mitigated by surgical excision. Improving post-operative testing could reduce reexcisions and improve long-term outcomes.

[0041] Clinical utility involves a balance between Type I and II errors. The clinical consequence of Type II errors (False Negatives) is that undetected positive margins place patients at high risk of recurrence. Some estimate that microscopy has a Type II error rate of 19% (patients who have positive margins but test negative). Assuming RNA Seq

performance is a reasonable indicator of clinical performance, a Type II error rate <5% represents a 75-100% improvement over existing methods. However, exclusive focus on Type II errors would be insufficient; high Type I errors (False Positives) would result in overtreatment. Surgeons may even avoid using a test with high Type I errors (False Positives) because it would trigger unnecessary reexcisions. Some estimates have placed Type I errors (False Positives) as 65% using existing microscopy methods. Reducing Type I errors from 65% to 5% would reduce unnecessary surgical reexcisions >90%.

III. Overview of Ductal Lavage

[0042] There is an urgent need to improve breast cancer screening and evaluation. Current screening tests have rates of false negative results, which fail to detect potentially lethal tumors. Current screening tests also have high rates of false positive results, which lead to invasive biopsies in patients who do not have breast cancer. Error rates of existing tests are not uniform. For example, it is not clear from current evidence whether the tradeoff is beneficial for screening mammography in women less than 50 years old. In the U.S., only 0.5% of women who are screened have cancer, but approximately 10% of women who undergo breast cancer screening require additional tests. On a population level, the false positive rate of breast cancer screening is therefore approximately 9.5%.

[0043] Mammography is the most widely used screening modality for the detection of breast cancer. There is conflicting evidence about whether screening mammography decreases breast cancer mortality. The evidence is strongest for women aged 50 to 69 years. However, screening in all age groups is also associated with harms. Harms can include unnecessary invasive procedures for patients who do not have breast cancer, and overdiagnosis, which is the detection of tumors that are not clinically significant. The error rates for mammography in women less than 50 are so high relative to the incidence of invasive breast cancers that the benefit of mammography is uncertain for women between 40 to 49 years old. In 2014, the Canadian National Breast Screening Study completed 25 years of follow-up and found no survival benefit associated with screening mammograms for women of all ages. While it is debatable how these findings should be applied to individual patients, it is clear that screening technologies are insufficient.

[0044] Alternative imaging technologies sometimes provide benefit for high-risk

populations, or as adjuncts to mammography, but are not recommended as primary screening tools for the general population. This group of technologies includes molecular breast imaging, ultrasound, and magnetic resonance imaging.

[0045] In the past, patients were advised to perform breast self-exams, but subsequent studies found that breast self-exams have no mortality benefit. Breast exams performed by clinicians (Clinical Breast Exams, CBE) have not been evaluated as an independent screening test. This leaves patients with poor options for early cancer detection, and limited options to determine whether a suspicious screening result warrants an invasive diagnostic procedure.

IV. Overview of a Molecular Test for Complete Response

[0046] Described herein is a method for analysis of residual tumor cells. The method and kits disclosed herein can identify improved treatment regimens. Accordingly, disclosed herein are post-operative devices and methods for obtaining and analyzing gene expression from cells from patient samples (e.g. from an excisional surgical biopsy) for residual disease. A panel of one to three cDNAs can serve as biomarkers to distinguish invasive breast cancer from adjacent healthy tissue with an accuracy of 96-100%. When cross-validated on 939 RNA Seq samples, the disclosed 3-gene test had a 96% Accuracy, 96% Sensitivity, and 94% Specificity. On an independent test set of 75 RNA Seq samples, the 3-gene test had a 97% Accuracy, 98% Sensitivity, 96% Specificity, 98% Positive Predictive Value, and 96 % Negative Predictive Value. We used The Cancer Genome Atlas (TCGA) project from the National Cancer Institute for biomarker discovery to identify a cohort of biomarkers from a population of subjects afflicted with a cancer. In contrast to many freely available datasets, the Biospecimen Core maintains rigorous protocols and quality controls that increase our confidence in pre-analytical variables. mRNA was profiled by RNA Seq (h=1,218) and microarray (n=l32). Subsets from the cohort of biomarkers were identified in subsequent analysis and informed a selection of biomarkers that correctly identified a cancer with high sensitivity and specificity.

V. Overview of a Molecular Test for Positive Surgical Margins

[0047] mRNAs are promising biomarkers because changes in cell and tissue morphology necessarily involve changes in gene activity and are therefore ideally situated to improve margin analysis. Moreover, we can now catalog tumor mRNAs across the genome. Finally, clinical labs routinely perform sensitive nucleic acid tests, positioning this qPCR assay for rapid adoption.

[0048] Prosigna® (PAM50 gene expression test) has 510K clearance from the FDA as a prognostic test for the risk of recurrence, in conjunction with clinical factors. However, by design, half of the 50 mRNAs in PAM50 are expressed at lower levels in tumors than in healthy tissues, and PAM50 is only valid when at least 50% of the sample is tumor. The PAM50 strategy of using genes that are downregulated in tumors could therefore not be used to detect rare tumor cells. Since our clinical indication involves detecting tumor cells in a population of healthy cells, we validated tumor-specific mRNAs with high expression in tumors.

[0049] Described herein is a method for analysis of residual tumor cells. The method and kits disclosed herein can identify complete excision of malignant tissue from patients.

Accordingly, disclosed herein are post-operative devices and methods for obtaining and analyzing gene expression from cells from patient samples (e.g. on the surface of surgical specimens) for residual disease. Nucleic acid tests for residual tumor cells provide a powerful solution to address positive surgical margins when combined with methods to acquire samples from the surface of a surgical sample.

VI. Overview of Molecular Test for Breast Cancer Screening

Described herein is a method for analysis of rare tumor cells. The method and kits disclosed herein can identify rare cancer cells, even when those tumor cells are not found in the context of healthy tissue. Accordingly, disclosed herein are screening devices and methods for obtaining and analyzing gene expression from cells from patient samples (e.g. nipple aspirates from ductal lavage) for disease. Disclosed herein are also adjuvant devices and methods to determine whether a screening test result warrants further investigation.

VII. Definitions

[0050] As used in the specification and in the claims, the singular form“a,”“an,” and“the” include plural referents unless the context clearly dictates otherwise.

[0051] The term“subject” or“patient” can include human or non-human animals. Thus, the methods and described herein are applicable to both human and veterinary disease and animal models. Preferred subjects are“patients,” e.g., living humans that are receiving medical care for a disease or condition (e.g., cancer). This includes persons with no defined illness who are being investigated for signs of pathology. The methods described herein are particularly useful for the evaluation of patients having or suspected of having breast adenocarcinomas. [0052] Biomarkers broadly refer to any characteristics that are objectively measured and evaluated as indicators of normal biological processes, pathogenic processes, or pharmacologic responses to therapeutic intervention. Unless otherwise noted, the term biomarker as used herein specifically refers to biomarkers that have biophysical properties, which allow their measurements in biological samples (e.g., plasma, serum, lavage, biopsy). Unless otherwise noted, the term biomarker is used interchangeably with“molecule biomarker” or“molecular markers.” Examples of biomarkers include nucleic acid biomarkers (e.g., oligonucleotides or polynucleotides), peptides or protein biomarkers, lipids, and lipopolysaccharide markers.

[0053] The term“polynucleotide” or“nucleic acid” as used herein refers to a polymeric form of nucleotides of any length, either ribonucleotides or deoxyribonucleotides, that comprise purine and/or pyrimidine bases, or other naturally modified nucleotide bases. Polynucleotides of the embodiments of the invention include sequences of deoxyribonucleic acid (DNA), ribonucleic acid (RNA), or DNA copies of ribonucleic acid (cDNA), all of which may be isolated from natural sources, recombinantly produced, or artificially synthesized. The polynucleotides and nucleic acids may exist as single-stranded or double-stranded.

[0054] The term“primer” as used herein refers to an oligonucleotide which is capable of acting as a point of initiation of synthesis when placed under conditions in which synthesis of a primer extension product, which is complementary to a nucleic acid strand, is induced, i.e., in the presence of nucleotides and an inducing agent such as a DNA polymerase and at a suitable temperature and pH. The primer may be either single-stranded or double- stranded and must be sufficiently long to prime the synthesis of the desired extension product in the presence of the inducing agent. The exact length of the primer will depend upon many factors, including temperature, source of primer and the method used. For example, for diagnostic applications, depending on the complexity of the target sequence, the oligonucleotide primer typically contains 15-35 or more nucleotides, although it may vary for certain biomarkers or

applications.

[0055]“Biological sample” as used herein is a sample of biological tissue or chemical fluid that is suspected of containing a biomarker or an analyte of interest. The sample may be an ex vivo sample or in vivo sample. Samples include, for example, tissue biopsies, e.g., from the breast or any other tissue suspected to be affected by, for instance, a metastasis of a cancer. The biopsy can be a liquid biopsy or a solid tissue biopsy. The sample can be a surgical excision from a tissue margin or another area suspected to be affected. A sample may be suspended or dissolved in, e.g., buffers, extractants, solvents, and the like. The terms sample and specimen can be used interchangeably herein. [0056] Ranges can be expressed herein as from“about” one particular value, and/or to “about” another particular value. When such a range is expressed, another embodiment includes from the one particular value and/or to the other particular value. Similarly, when values are expressed as approximations, by use of the antecedent“about,” it will be understood that the particular value forms another embodiment. It will be further understood that the endpoints of each of the ranges are significant both in relation to the other endpoint, and independently of the other endpoint. The term“about” as used herein refers to a range that is 15% plus or minus from a stated numerical value within the context of the particular usage. For example, about 10 would include a range from 8.5 to 11.5.

VIII. Samples

[0057] Methods for detecting molecules (e.g., nucleic acids, proteins, etc.) in a subject in order to detect, diagnose, monitor, or evaluate the presence of residual cancer are described in this disclosure. In some cases, the molecules are circulating molecules. In some cases, the molecules are expressed in the cytoplasm of blood, endothelial, or organ cells. In some cases, the molecules are expressed on the surface of blood, endothelial, or organ cells.

[0058] The methods, kits, and systems disclosed herein can be used to classify one or more samples from one or more subjects. A sample can be any material containing tissues, cells, nucleic acids, genes, gene fragments, expression products, polypeptides, exosomes, gene expression products, or gene expression product fragments of a subject to be tested. A sample can include but is not limited to, tissue, cells, or biological material from cells or derived from cells of an individual. The sample can be a heterogeneous or homogeneous population of cells or tissues. The sample can be a fluid that is acellular or depleted of cells (e.g., serum). In some cases, the sample is from a single patient. In some cases, the method comprises analyzing multiple samples at once, e.g., via massively parallel multiplex expression analysis on protein arrays or the like.

[0059] The sample may be obtained using any suitable method. The sample may be obtained by a minimally-invasive method, e.g., venipuncture or ductal lavage. The sample obtained by venipuncture may comprise whole blood or a component thereof (e.g. serum, white blood cells). Ductal lavage may be performed by e.g. the method described in US20020058887A1, which is incorporated by reference herein. Alternatively, the sample may be obtained an invasive method, such as by biopsy. Biopsies could include core biopsies, punch biopsies, incisional biopsies and excisional biopsies. A sample obtained by surgical excision may comprise a subsection of an excised tissue chunk (e.g. a representative cross-section of tissue). A sample obtained by surgical excision may comprise a cell-dissociated or homogenized chunk of some or all of the excised tissue. A sample obtained by surgical excision may comprise a surface sample of excised tissue. A surface sample of excised tissue may comprise a“touch prep” sample which reflects the population of cells along the margins of the excised tissue ( e.g . tumor).

[0060] In some embodiments, obtaining a sample comprises directly isolating a sample from a patient. In some embodiments, obtaining a sample comprises obtaining a sample previously isolated from a patient. In some embodiments, obtaining a sample comprises obtaining polynucleotides isolated from a sample previously isolated from a patient.

[0061] The cellular specimen may be obtained using imprint cytology acquisition strategies, one form of which is a‘touch prep’ or similar method. A‘touch prep’ is known as a type of imprint cytology. Generally, the term‘touch prep’ refers to both the process of preparing the slide, rapid staining the slide, and analyzing the slide under a microscope. The‘touch prep’ method may involve smearing or spreading the obtained cellular specimen onto a slide or a plurality of slides. The‘touch prep’ method may involve pressing the slide to the biological sample. The‘touch prep’ method may involve pressing the slide to the excised tissue. The ‘touch prep’ method may involve pressing the slide to a tissue on or within the subject. The ‘touch prep’ method may involve pressing the slide to an area, wall or margin surrounding a tissue or biological sample on or within the subject. The‘touch prep’ method may involve pressing the slide to an area, wall or margin surrounding a site where a tissue was excised. Touch prep may be performed in, e.g. less than about 60 minutes, less than about 55 minutes, less than about 50 minutes, less than about 45 minutes, about less than 40 minutes, about less than 35 minutes, about less than 30 minutes, about less than 25 minutes, less than about 20 minutes, less than about 15 minutes, less than about 10 minutes, less than about 5 minutes, less than about 3 minutes, less than about 2 minutes, less than about 1 minute, less than about 30 seconds, less than about 10 seconds, less than about 5 seconds, less than about 2 seconds, or less than about 1 second. The‘touch prep’ method may be performed in a few seconds per slide. The‘touch prep’ method may be performed by a surgeon, a nurse, an assistant, a cytopathologist, a person with no medical training or the subject. The‘touch prep’ method may be operated manually. The‘touch prep’ method may be operated automatically by a machine. The‘touch prep’ method may be performed intraoperatively to detect or rule out malignant cells along the surgical margin (e.g. during a breast lumpectomy). During the ‘touch prep’ method, the excised tissue may be pressed against a sample collection unit which is a glass slide coated with poly-Lysine, or other surface. The cellular specimen obtained by a touch prep method may be used to determine the presence or absence of malignant cells along the margin of excised tissue. In some cases, the surface comprises sample collection unit. In some cases, the sample is then applied to a sample input unit of a device. In some embodiments, the touch prep sample may be obtained according to the methodology described in US20040030263A1, which is incorporated by reference herein.

[0062] In some embodiment, the samples comprise tissue samples and are prepared by tumor dissociation/homogenization. In some embodiments, this is accomplished using the Miltenyi Biotec Tumor Dissociation Kit in combination with a gentleMACS Tissue Dissociator to homogenize tissue samples in a sterile environment. The Tissue Dissociator uses disposable Miltenyi M tubes with rotor-stators that are built into the tube lids. Frozen samples may be used to achieve more comsistent yields. Tissue is added to cell lysis in buffer directly in the disruptor tube. After dissociation and lysis, RNA is isolated using an RNA isolation kit, such as Qiagen RNeasy Mini Kit. This method can isolate high-quality RNA from both tumor and adipose-based tissues. Larger specimens may be divided into smaller pieces depending on maximum tissue input. If tissue dissociation alone does not collect enough high-quality RNA for RTqPCR, samples may be pre-incubated with enzymatic treatments (e.g. Collagenases). Enzymatic treatments may be applied during mechanical dissociation, which others have validated for the GentleMACS Tissue Dissociator.

[0063] In some embodiments, the methods or compositions herein are capable of detecting breast cancer in a sample from a cancer patient, detecting residual breast cancer in a sample from a cancer patient (e.g. post-chemotherapy/radiation/surgery) , or distinguishing between breast cancer and surrounding healthy breast tissue. In some embodiments, the detection is based on a minimal amount of polynucleotides or nucleic acids isolated from a sample.

[0064] In some embodiments the minimal amount of polynucleotides or nucleic acids isolated from the sample is at least 10 ng, 50 ng, lOOng, 200 ng, 500ng, 1 mg, 2 mg, 3 mg, 4 mg, 5 mg, 10 mg, 15 mg, 20 mg, 50 mg, or 500ng. In some embodiments, the methods or compositions herein are capable of detecting residual cancer in a sample from a patient, or distinguishing between cancer and surrounding healthy tissue, based on a minimal weight of tissue sample used to isolate polynucleotides or nucleic acids. In some embodiments, the minimal amount of tissue sample is at least 100 ng, 200 ng, 500 ng, lmg, 2 mg, 3 mg, 4 mg, 5 mg, 10 mg, 15 mg, 20 mg, 50 mg, 100 mg, 200 mg, 300 mg, or 500 mg.

IX. Biomarkers

[0065] The term“biomarker” as used herein refers to a measurable indicator of some biological state or condition. In some instances, a biomarker can be a substance found in a subject, a quantity of the substance, or some other indicator. For example, a biomarker can be the amount of a protein and/or other gene expression products in a sample. In some embodiments, a biomarker is a total level of protein in a sample. In some embodiments, a biomarker is a total level of a particular type of nucleic acid ( e.g . RNA, cDNA) in a sample. In some embodiments, a biomarker is a therapeutic target, or an indicator of response to therapy.

[0066] The methods, compositions and systems as described here also relate to the use of biomarker panels for purposes of research, identification, diagnosis, classification, treatment or to otherwise characterize the status of cancer in a patient. Sets of biomarkers useful for classifying biological samples are provided, as well as methods of obtaining such sets of biomarkers. Often, the pattern of levels of biomarkers in a panel (also known as a signature) is determined from a control sample or population and then used to evaluate the signature of the same panel of biomarkers in an experimental sample or population, such as by a measure of similarity between the sample signature and the reference signature.

[0067] In some embodiments, the panels of biomarkers described herein are useful for the detection of breast cancer (e.g. detection of positive surgical margins on a biopsy sample, detection of residual disease in a cancer patient post-radiation/chemotherapy/surgery, or detection of disease in a patient suspected of having cancer). In some embodiments the breast cancer is invasive adenocarcinoma, invasive ductal breast cancer, invasive lobular breast cancer, or a combination thereof. In some embodiments, the breast cancer is HER2 positive, ER (estrogen receptor) positive, or PR (progesterone receptor) positive, or a combination thereof. In some embodiments, the breast cancer is HER2 negative, ER

(estrogen receptor) negative, or PR (progesterone receptor) negative, or a combination thereof.

[0068] In some embodiments, the methods herein comprise measuring expression levels of genes selected from the group consisting essentially of Matrix Metallopeptidase 11

( MMP11 ), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain

( COL10A1 ). In some embodiments, the methods herein comprise measuring expression levels of genes selected from the group consisting of Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain (COLIOAI). In some embodiments the methods herein comprise measuring expression levels of genes selected from the group consisting essentially of Matrix Metallopeptidase 11 ( MMP11 ) and integrin binding sialoprotein (IBSP). In some embodiments the methods herein comprise measuring expression levels of genes selected from the group consisting of Matrix Metallopeptidase 11 (MMP11) and integrin binding sialoprotein (IBSP). [0069] The biomarkers that form the basis for the 3-gene test described herein ( MMP11 , IBSP, and COL10A1 ) particularly useful in that their expression is higher (upregulated) in cancerous tissues than in normal tissues. As a result, the fraction of a sample that must contain cancerous cells for the sample to be labeled as positive is much lower than for a test that depends on genes that have decreased expression (downregulated) in cancerous tissue.

In some embodiments, the methods, compositions and systems as described here also relate to the use of a biomarker test of research, identification, diagnosis, classification, treatment or to otherwise characterize the status of cancer in a patient, wherein at least one of Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain ( COL10A1 ) are higher in said cancer than in healthy tissue. In some embodiments, at least two of Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain (COLIOAI) are higher in said cancer than in healthy tissue. In some embodiments, the levels of each of Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain (COLIOAI) are higher in said cancer than in healthy tissue.

X. Biomarker Expression Profiles

[0070] The methods, kits, and systems disclosed herein may comprise specifically detecting, profiling, or quantitating biomolecules (e.g., nucleic acids, DNA, RNA, polypeptides, etc.) that are within the biological samples to determine an expression profile. In some instances, genomic expression products, including RNA, or polypeptides, may be isolated from the biological samples. In some cases, nucleic acids, DNA, RNA, polypeptides may be isolated from a cell-free source. In some cases, nucleic acids, DNA, RNA, polypeptides may be isolated from cells derived from the cancer patient. In some cases, the molecules detected are derived from molecules endogenously present in the sample via an enzymatic process (e.g, cDNA derived from reverse transcription of RNA from the biological sample followed by amplification).

[0071] Expression profiles are preferably measured at the nucleic acid level, meaning that levels of mRNA or nucleic acid derived therefrom (e.g, cDNA or RNA) are measured. An expression profile refers to the expression levels of a plurality of genes in a sample. A nucleic acid derived from mRNA means a nucleic acid synthesized using mRNA as a template. Methods of isolation and amplification of mRNA are described in, e.g, Chapter 3 of Laboratory Techniques in Biochemistry and Molecular Biology: Hybridization With Nucleic Acid Probes, Part I. Theory and Nucleic Acid Preparation, (P. Tijssen, ed.) Elsevier, N.Y. (1993). If mRNA or a nucleic acid therefrom is amplified, the amplification is performed under conditions that approximately preserve the relative proportions of mRNA in the original samples, such that the levels of the amplified nucleic acids can be used to establish phenotypic associations representative of the mRNAs.

[0072] In some embodiments, expression levels are determined by direct detection of nucleic acids. Such methods include e.g. gel or capillary electrophoresis, wherein specifically amplified DNA is detected by its intrinsic fluorescence/absorbance, or by complexing with a suitable absorbent or fluorescent DNA-binding dye. Such methods can be used alongside PCR or RT-PCR with forward and reverse primers against specific genes to detect levels of genes within nucleic acids isolated from a sample.

[0073] In other methods, expression levels are determined by NanoString™ assay.

NanoString™ based assays are described in the U.S. Patent Nos. 8,415, 102, 8,519,115, and 7,919,237, which are herein incorporated by reference in their entirety. NanoString's NCOUNTER technology is a variation on the DNA microarray. It uses molecular "barcodes" and microscopic imaging to detect and count up to several hundred unique transcripts in one hybridization reaction. Each color-coded barcode is attached to a single target-specific probe corresponding to a target of interest. The protocol typically includes hybridization

(employing two ~50 base probes per mRNA that hybridize in solution; the reporter probe carries the signal, while the capture probe allows the complex to be immobilized for data collection); purification and immobilization (after hybridization, the excess probes are removed and the probe/target complexes are aligned and immobilized in the cartridge); and data collection (sample cartridges are placed in a digital analyzer instrument for data collection; color codes on the surface of the cartridge are counted and tabulated for each target molecule). The protocol is carried out with a prep station, which is an automated fluidic instrument that immobilizes code set complexes for data collection, and a digital analyzer, which derives data by counting fluorescent barcodes. Code set complexes are custom-made or pre-designed sets of color-coded probes pre -mixed with a set of system controls. Probes for the barcode-based assay can be designed according to desired variables such as melting temperature (Tm) and specificity for the template mRNA/cDNA to be detected.

[0074] In other methods, expression levels are determined by so-called“real time amplification” methods also known as quantitative PCR (qPCR) or Taqman. The basis for this method of monitoring the formation of amplification product formed during a PCR reaction with a template using oligonucleotide probes/oligos specific for a region of the template to be detected. In some embodiments, qPCR or Taqman are used immediately following a reverse-transcriptase reaction performed on isolated cellular mRNA; this variety serves to quantitate the levels of individual mRNAs during qPCR.

[0075] Taqman uses a dual-labeled fluorogenic oligonucleotide probe. The dual labeled fluorogenic probe used in such assays is typically a short (ca. 20-25 bases) polynucleotide that is labeled with two different fluorescent dyes. The 5’ terminus of the probe is typically attached to a reporter dye and the 3’ terminus is attached to a quenching dye. Regardless of labelling or not, the qPCR probe is designed to have at least substantial sequence

complementarity with a site on the target mRNA or nucleic acid derived from. Upstream and downstream PCR primers that bind to flanking regions of the locus are also added to the reaction mixture. When the probe is intact, energy transfer between the two fluorophores occurs and the quencher quenches emission from the reporter. During the extension phase of PCR, the probe is cleaved by the 5’ nuclease activity of a nucleic acid polymerase such as Taq polymerase, thereby releasing the reporter from the polynucleotide-quencher and resulting in an increase of reporter emission intensity which can be measured by an appropriate detector. The recorded values can then be used to calculate the increase in normalized reporter emission intensity on a continuous basis and ultimately quantify the amount of the mRNA being amplified. mRNA levels can also be measured without amplification by hybridization to a probe, for example, using a branched nucleic acid probe, such as a QuantiGene® Reagent System from Panomics. This format of test is particularly useful for the multiplex detection of multiple genes from a single sample reaction, as each fluorophore/quencher pair attached to an individual probe may be spectrally orthogonal to the other probes used in the reaction such that multiple probes (each directed against a different gene product) can be detected during the amplification/detection reaction.

[0076] qPCR can also be performed without a dual-labeled fluorogenic probe by using a fluorescent dye (e.g. SYBR Green) specific for dsDNA that reflects the accumulation of dsDNA amplified specific upstream and downstream oligonucleotide primers. The increase in fluorescence during the amplification reaction is followed on a continuous basis and can be used to quantify the amount of mRNA being amplified.

[0077] For qPCR or Taqman, the levels of particular genes may be expressed relative to one or more internal control gene measured from the same sample using the same detection methodology. Internal control genes may include so-called“housekeeping” genes (e.g. ACTB, B2M, UBC, GAPD and HPRT1). In some embodiments, the one or more internal control gene is TTC5, C2orf44, or Chr3. [0078] In some embodiments, for qPCR or Taqman detection, a“pre-amplification” step is performed on cDNA transcribed from cellular RNA prior to the quantitatively monitored PCR reaction. This serves to increase signal in conditions where the natural level of the RNA/cDNA to be detected is very low. Suitable methods for pre-amplification include but are not limited LM-PCR, PCR with random oligonucleotide primers (e.g. random hexamer PCR), PCR with poly-A specific primers, and any combination thereof.

[0079] In some embodiments, for qPCR or Taqman detection, an RT-PCR step is first performed to generate cDNA from cellular RNA. Such amplification by RT-PCR can either be general (e.g. amplification with partially/fully degenerate oligonucleotide primers) or targeted (e.g. amplification with oligonucleotide primers directed against specific genes which are to be analyzed at a later step).

[0080] In other methods, expression levels are determined by sequencing, such as by RNA sequencing or by DNA sequencing (e.g., of cDNA generated from reverse-transcribing RNA (e.g, mRNA) from a sample). Sequencing may also be general (e.g. with amplification using partially/fully degenerate oligonucleotide primers) or targeted (e.g. with amplification using oligonucleotide primers directed against specific genes which are to be analyzed at a later step). Sequencing may be performed by any available method or technique. Sequencing methods may include: Next Generation sequencing, high-throughput sequencing, pyrosequencing, classic Sanger sequencing methods, sequencing-by-ligation, sequencing by synthesis, sequencing-by-hybridization, RNA-Seq (Illumina), Digital Gene Expression (Helicos), next generation sequencing, single molecule sequencing by synthesis (SMSS) (Helicos), Ion Torrent Sequencing Machine (Life Technologies/Thermo-Fisher), massively- parallel sequencing, clonal single molecule Array (Solexa), shotgun sequencing, Maxim- Gilbert sequencing, primer walking, and any other sequencing methods known in the art.

[0081] Measuring gene expression levels may comprise reverse transcribing RNA (e.g., mRNA) within a sample in order to produce cDNA. The cDNA may then be measured using any of the methods described herein (e.g, qPCR, sequencing, etc.).

[0082] Alternatively, or additionally, expression levels of genes can be determined at the protein level, meaning that levels of proteins encoded by the genes discussed above are measured. Several methods and devices are well known for determining levels of proteins including immunoassays such as sandwich, competitive, or non-competitive assay formats, to generate a signal that is related to the presence or amount of a protein analyte of interest. Immunoassays such as, but not limited to, lateral flow, enzyme-linked immunoassays (ELISA), radioimmunoassays (RIAs), and competitive binding assays may be utilized. Numerous formats for antibody arrays have been described proposed employing antibodies. Other ligands having specificity for a particular protein target can also be used, such as synthetic antibodies.

XI: Sensitivity, Specificity, Accuracy and other measures of performance

[0083] The methods provided herein can detect the presence of residual disease, such as a positive margin on a surgical cancer biopsy or presence of disease (e.g. of in a sample from a cancer patient with a high degree of accuracy, sensitivity, and/or specificity. In some cases, the accuracy (e.g., for detecting residual disease, or distinguishing between residual disease and surrounding healthy tissue) is at least about 75%, 80%, 85%, 90%, 91%, 92%, 93%,

94%, 95%, 96%, 97% 98% or 99%. In some cases, the sensitivity (e.g., for detecting residual disease, or distinguishing between residual disease and surrounding healthy tissue) is at least about 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97% 98% or 99%. In some cases, the specificity (e.g, for detecting residual disease, or distinguishing between residual disease and surrounding healthy tissue) is at least about 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97% 98% or 99%.. In some cases, the positive predictive value (e.g, for detecting residual disease, or distinguishing between residual disease and surrounding healthy tissue) of the method at least 75%, 80%, 85%, 90%, 91%, 92%, 93%, 94%, 95%,

96%, 97% 98% or 99%.. The AUC after thresholding in any of the methods provided herein may be at least about 0.7, 0.75, 0.8, 0.85, 0.9, 0.91, 0.92, 0.93, 0.94, 0.95. 0.96, 0.97, 0.98, 0.99, 0.995, or 0.999. In some embodiments, the methods disclosed herein have a positive predictive value of at least about 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%. In some embodiments, the methods disclosed herein have a negative predictive value of at least about 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or 99%.

XII. Clinical Applications

[0084] The methods, compositions, systems and kits provided herein can be used to detect, diagnose, predict or monitor a condition of a pregnant patient. In some instances, the methods, compositions, systems and kits described herein provide information to a medical practitioner that can be useful in making a therapeutic decision. Therapeutic decisions can include decisions to: continue with a particular therapy, modify a particular therapy, alter the dosage of a particular therapy, stop or terminate a particular therapy, altering the frequency of a therapy, introduce a new therapy, introduce a new therapy to be used in combination with a current therapy, or any combination of the above. In some cases, the methods provided herein can be applied in an experimental setting, e.g., a clinical trial. In some embodiments, the guidance of a test result herein (e.g. presence of residual disease) may be used to determine the end of a course of therapy ( e.g . standard chemotherapy regimens). In some embodiments, the guidance of a test result herein (e.g. presence of residual disease) may be used to indicate the location of a further tumor excision to be performed on the patient (e.g. in the case where the test is used in combination with touch prep multiple touch prep samples derived as described above to indicate where surgical margins have been insufficient in an excised sample).

XIII. Monitoring a Condition of a Patient

[0085] Provided herein are methods, systems, kits and compositions for monitoring a condition of a cancer patient (e.g. presence of residual disease). Often, the monitoring is conducted by serial testing, such as serial non-invasive tests, serial minimally-invasive tests (e.g, blood draws, ductal lavage), or some combination thereof.

[0086] In some instances, the cancer patient is monitored as needed using the methods described herein. Alternatively the cancer patient can be monitored weekly, monthly, or at any pre-specified intervals. In some instances, the cancer patient is monitored at least once every 24 hours. In some instances the cancer patient is monitored at least once every 1 day to 30 days. In some instances the cancer patient is monitored at least once every at least 1 day.

In some instances the cancer patient is monitored at least once every at most 30 days. In some instances the cancer patient is monitored at least once every 1 day to 5 days, 1 day to 10 days, 1 day to 15 days, 1 day to 20 days, 1 day to 25 days, 1 day to 30 days, 5 days to 10 days, 5 days to 15 days, 5 days to 20 days, 5 days to 25 days, 5 days to 30 days, 10 days to 15 days,

10 days to 20 days, 10 days to 25 days, 10 days to 30 days, 15 days to 20 days, 15 days to 25 days, 15 days to 30 days, 20 days to 25 days, 20 days to 30 days, or 25 days to 30 days. In some instances the cancer patient is monitored at least once every 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11,12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 28, 29, 30 or 31 days. In some instances, the cancer patient is monitored at least once every 1, 2, 3, or 6 months.

XIV. Sequences and Embodiments of Combinations of Sequences

[0087] The primers disclosed herein, such as a pair of primers as described herein, specifically a forward primer (“F”) and a reverse primer (“R”) for both strands to be detected, can be in a composition in amounts effective to permit detection of native, mutant, reference, or control sequences. Detection of native, mutant, reference, or control sequences is accomplished using any of the methods described herein or known by one of ordinary skill in the art in the art for detecting a specific nucleic acid molecule in a sample. The primers disclosed herein may be provided as part of a kit. A kit can also comprise buffers, nucleotide bases and other compositions to be used in hybridization and/or amplification reactions. In other cases, the primers described herein may be part of a device.

[0088] In some embodiments a panel of nucleic acids is detected in a sample from a patient. A panel of one to three cDNAs can serve as biomarkers to distinguish invasive breast cancer from adjacent healthy tissue. A panel of one to three cDNAs can serve to residual breast cancer post-chemotherapy, post-radiation treatment, or post-surgical excision of tumor(s). Such cDNA panels may comprise IBSP, MMP11, and/or COL10A1 cDNA. A panel may comprise two or three genes selected from IBSP, MMP11, and COL10A1, which can be amplified using the primers disclosed herein. In some embodiments, the relative levels of cDNA panels may be assessed relative to the cDNA levels of a reference gene panel. Such reference gene panel may comprise TTC5 and/or C2orf44, which can be amplified using the primers disclosed herein. In some cases, the single genes or gene panels are compared to a negative control for genomic DNA, for example, chr3 gDNA, which can be amplified using the primers disclosed herein.

XV. Primers

[0089] Exemplary forward primers for IBSP cDNA

[0090] Exemplary reverse primers for IBSP cDNA

[0091] Exemplary forward primers for L/L/E / / cDNA

[0092] Exemplary reverse primers for A7A7/V / cDNA

[0093] Exemplary forward primers for COL10A1 cDNA

[0094] Exemplary reverse primers for COL10A1 cDNA

[0095] Exemplary forward primers for TTC5 cDNA

[0096] Exemplary reverse primers for TTC5 cDNA

[0097] Exemplary forward primers for C2orf44 cDNA

[0098] Exemplary reverse primers for C2orf44 cDNA

[0099] Exemplary forward primers for Chr3 gDNA

[00100] Exemplary reverse primers for chr3 gDNA

XVI. Exemplary Forward/Reverse primer combinations for RTqPCR measurement of IBSPcDNA

[00101]

[00102] Exemplary Forward/Reverse primer combinations for RTqPCR/qPCR measurement oϊMMRII cDNA

[00103] Exemplary Forward/Reverse primer combinations for RTqPCR/qPCR measurement of COL10A1 cDNA

[00104] Exemplary Forward/Reverse Primer Combinations for RTqPCR/qPCR measurement of TTC5 cDNA

[00105] Exemplary Forward/Reverse Primer Combinations for RTqPCR/qPCR measurement of C2orf44 cDNA

[00106] Exemplary Forward/Reverse Primer Combinations for

RT qPCR/qPCR/RTPCR of Chr3 gDNA

[00107] In some embodiments, the nucleic acids disclosed herein may be used a biomarker. For example, a portion of the cDNA sequence of MMP I /, IBSP, or COL10A1 may be used as a biomarker to detect cancer.

[00108] In some embodiments, the sequence of an MMP 11 cDNA is according to:

[00109] ATAAGGGGCGGCGGCCCGGAGCGGCCCAGCAAGCCCAGCAGCCCC

GGGGCGGATGGCTCCGGCCGCCTGGCTCCGCAGCGCGGCCGCGCGCGCCCTCCT

GCCCCCGATGCTGCTGCTGCTGCTCCAGCCGCCGCCGCTGCTGGCCCGGGCTCTG

CCGCCGGACGCCCACCACCTCCATGCCGAGAGGAGGGGGCCACAGCCCTGGCAT

GCAGCCCTGCCCAGTAGCCCGGCACCTGCCCCTGCCACGCAGGAAGCCCCCCGG CCTGCCAGCAGCCTCAGGCCTCCCCGCTGTGGCGTGCCCGACCCATCTGATGGGC

TGAGTGCCCGCAACCGACAGAAGAGGTTCGTGCTTTCTGGCGGGCGCTGGGAGA

AGACGGACCTCACCTACAGGATCCTTCGGTTCCCATGGCAGTTGGTGCAGGAGCA

GGTGCGGCAGACGATGGCAGAGGCCCTAAAGGTATGGAGCGATGTGACGCCACT

CACCTTTACTGAGGTGCACGAGGGCCGTGCTGACATCATGATCGACTTCGCCAGG

TACTGGCATGGGGACGACCTGCCGTTTGATGGGCCTGGGGGCATCCTGGCCCATG

CCTTCTTCCCCAAGACTCACCGAGAAGGGGATGTCCACTTCGACTATGATGAGAC

CTGGACT ATCGGGGAT GACC AGGGC AC AGACCTGCTGC AGGT GGC AGCCC AT GA

ATTTGGCCACGTGCTGGGGCTGCAGCACACAACAGCAGCCAAGGCCCTGATGTC

CGCCTTCTACACCTTTCGCTACCCACTGAGTCTCAGCCCAGATGACTGCAGGGGC

GTTCAACACCTATATGGCCAGCCCTGGCCCACTGTCACCTCCAGGACCCCAGCCC

TGGGCCCCCAGGCTGGGATAGACACCAATGAGATTGCACCGCTGGAGCCAGACG

CCCCGCCAGATGCCTGTGAGGCCTCCTTTGACGCGGTCTCCACCATCCGAGGCGA

GCTCTTTTTCTTCAAAGCGGGCTTTGTGTGGCGCCTCCGTGGGGGCCAGCTGCAG

CCCGGCTACCCAGCATTGGCCTCTCGCCACTGGCAGGGACTGCCCAGCCCTGTGG

ACGCTGCCTTCGAGGATGCCCAGGGCCACATTTGGTTCTTCCAAGGTGCTCAGTA

CTGGGTGTACGACGGTGAAAAGCCAGTCCTGGGCCCCGCACCCCTCACCGAGCT

GGGCCTGGTGAGGTTCCCGGTCCATGCTGCCTTGGTCTGGGGTCCCGAGAAGAAC

AAGATCTACTTCTTCCGAGGCAGGGACTACTGGCGTTTCCACCCCAGCACCCGGC

GTGTAGACAGTCCCGTGCCCCGCAGGGCCACTGACTGGAGAGGGGTGCCCTCTG

AGATCGACGCTGCCTTCCAGGATGCTGATGGCTATGCCTACTTCCTGCGCGGCCG

CCTCTACTGGAAGTTTGACCCTGTGAAGGTGAAGGCTCTGGAAGGCTTCCCCCGT

CTCGTGGGTCCTGACTTCTTTGGCTGTGCCGAGCCTGCCAACACTTTCCTCTGACC

ATGGCTTGGATGCCCTCAGGGGTGCTGACCCCTGCCAGGCCACGAATATCAGGCT

AGAGACCCATGGCCATCTTTGTGGCTGTGGGCACCAGGCATGGGACTGAGCCCA

TGTCTCC T C AGGGGGAT GGGGT GGGGT AC A AC C AC CAT G AC A AC T GC CGGG AGG

GCCACGCAGGTCGTGGTCACCTGCCAGCGACTGTCTCAGACTGGGCAGGGAGGC

TTTGGCATGACTTAAGAGGAAGGGCAGTCTTGGGCCCGCTATGCAGGTCCTGGCA

AACCTGGCTGCCCTGTCTCCATCCCTGTCCCTCAGGGTAGCACCATGGCAGGACT

GGGGGAACTGGAGTGTCCTTGCTGTATCCCTGTTGTGAGGTTCCTTCCAGGGGCT

GGCACTGAAGCAAGGGTGCTGGGGCCCCATGGCCTTCAGCCCTGGCTGAGCAAC

TGGGCTGTAGGGCAGGGCCACTTCCTGAGGTCAGGTCTTGGTAGGTGCCTGCATC

TGTCTGCCTTCTGGCTGACAATCCTGGAAATCTGTTCTCCAGAATCCAGGCCAAA

AAGTTCACAGTCAAATGGGGAGGGGTATTCTTCATGCAGGAGACCCCAGGCCCT GGAGGCTGCAACATACCTCAATCCTGTCCCAGGCCGGATCCTCCTGAAGCCCTTT TCGCAGCACTGCTATCCTCCAAAGCCATTGTAAATGTGTGTACAGTGTGTATAAA CCTTCTTCTTCTTTTTTTTTTTTTAAACTGAGGATTGTCATTAAACACAGTTGTTTT CT AAAAAAAA AAAAAAAA (SEQ ID NO: 357)

[00110] In some embodiments, the sequence of an IBSP cDNA is according to:

[00111] GAGT GAGT GAGAGGGC AGAGGAAAT ACTC AATCTGT GCC ACTC AC

TGCCTTGAGCCTGCTTCCTCACTCCAGGACTGCCAGAGGAAGCAATCACCAAAAT

GAAGACTGCTTTAATTTTGCTCAGCATTTTGGGAATGGCCTGTGCTTTCTCAATGA

AAAATTT GC ATCGA AGAGTC AAAAT AGAGGATTCTGAAGAAAAT GGGGTCTTT A

AGTACAGGCCACGATATTATCTTTACAAGCATGCCTACTTTTATCCTCATTTAAAA

CGATTTCC AGTT C AGGGC AGT AGTGACTC ATCCGA AGAAAAT GGAGAT GAC AGT

T C AGAAGAGGAGGAGGAAGAAGAGGAGACTT C AAATGAAGGAGAAAAC AAT GA

AGAATCGAATGAAGATGAAGACTCTGAGGCTGAGAATACCACACTTTCTGCTAC

AACACTGGGCTATGGAGAGGACGCCACGCCTGGCACAGGGTATACAGGGTTAGC

TGCAATCCAGCTTCCCAAGAAGGCTGGGGATATAACAAATAAAGCTACAAAAGA

GA AGGA A AGT GAT GA AGA AGA AGAGGAGGA AGAGGA AGGA A AT GA A A AC GA A

GAAAGCGAAGCAGAAGTGGATGAAAACGAACAAGGCATAAACGGCACCAGTAC

CAACAGCACAGAGGCAGAAAACGGCAACGGCAGCAGCGGAGGAGACAATGGAG

A AGA AGGGGA AGA AGA A AGT GT C AC T GG AGCC A AT GC AGA AGAC ACC AC AGAG

ACCGGAAGGC AGGGC AAGGGCACCTCGAAGACAACAACCTCTCCAAATGGTGG

GTTTGAACCTACAACCCCACCACAAGTCTATAGAACCACTTCCCCACCTTTTGGG

AAAACCACCACCGTTGAATACGAGGGGGAGTACGAATACACGGGCGCCAATGAA

T ACGAC A ATGGAT AT GAAATCT AT GAAAGT GAGAACGGGGAACCTCGT GGGGAC

AATTACCGAGCCTATGAAGATGAGTACAGCTACTTTAAAGGACAAGGCTACGAT

GGCTATGATGGTCAGAATTACTACCACCACCAGTGAAGCTCCAGCCTGGGATGA

ATTCATCCATTCTGGCTTTGCATCCGGCTACCATTTTCGAAGTTCAACTCAGGAAG

GTGCAATATAACAAATGTGCATATTATAATGAGGAATGGTACTACCGTTCCAGAT

TTTCTGTAATTGCTTCTGCAAAGTAATAGGCTTCTTGTCCCTTTTTTTTCTGGCATG

TTATGGAATGATCATTGTAAATCAGGACCATTTATCAAGCAGTACACCAACTCAT

AAGATCAAATTTCATTGAATGGTTTGAGGTTGTAGCTCTATAAATAGTAGTTTTT

AAC AT GCCTGT AGT ATT GCT AACTGC A AAAAC AT ACTCTTT GT AC AAGAAGT GCT

TCTAAGAATTTCATTGACATTAATGACACTGTATACAATAAATGTGTAGTTTCTTA

ATCGCACTACCTATGCAACACTGTGTATTAGGTTTATCATCCTCATGTATTTTTAT GTGACCTGTATGTATATTCTAATCTACGAGTTTTATCACAAATAAAAATGCAATC CTTCAAATGTGTTATAATTAAAAAA (SEQ ID NO:358)

[00112] In some embodiments, the sequence of a COL10A1 cDNA is according to:

[00113] C AC CUU CU GC ACU GCUC AU CU GGGC AGAGGA AGCUUC AGA A AGC

UGCCAAGGCACCAUCUCCAGGAACUCCCAGCACGCAGAAUCCAUCUGAGAAUA

UGCUGCCACAAAUACCCUUUUUGCUGCUAGUAUCCUUGAACUUGGUUCAUGGA

GUGUUUUACGCUGAACGAUACCAAAUGCCCACAGGCAUAAAAGGCCCACUACC

CAACACCAAGACACAGUUCUUCAUUCCCUACACCAUAAAGAGUAAAGGUAUAG

CAGUAAGAGGAGAGCAAGGUACUCCUGGUCCACCAGGCCCUGCUGGACCUCGA

GGGCACCCAGGUCCUUCUGGACCACCAGGAAAACCAGGCUACGGAAGUCCUGG

ACUCCAAGGAGAGCCAGGGUUGCCAGGACCACCGGGACCAUCAGCUGUAGGGA

AACCAGGUGUGCCAGGACUCCCAGGAAAACCAGGAGAGAGAGGACCAUAUGG

ACCAAAAGGAGAUGUUGGACCAGCUGGCCUACCAGGACCCCGGGGCCCACCAG

GACCACCUGGAAUCCCUGGACCGGCUGGAAUUUCUGUGCCAGGAAAACCUGGA

CAACAGGGACCCACAGGAGCCCCAGGACCCAGGGGCUUUCCUGGAGAAAAGGG

U GC AC C AGG AGU C C CU GGU AU G A AU GG AC AG A A AGGGG A A AU GGG AU AU GGU

GCUCCUGGUCGUCCAGGUGAGAGGGGUCUUCCAGGCCCUCAGGGUCCCACAGG

ACCAUCUGGCCCUCCUGGAGUGGGAAAAAGAGGUGAAAAUGGGGUUCCAGGA

C AGC C AGGC AU C A A AGGU GAU AGAGGUUUU C CGGGAGA A AU GGG AC C A AUU G

GCCCACCAGGUCCCCAAGGCCCUCCUGGGGAACGAGGGCCAGAAGGCAUUGGA

AAGCCAGGAGCUGCUGGAGCCCCAGGCCAGCCAGGGAUUCCAGGAACAAAAGG

UCUCCCUGGGGCUCCAGGAAUAGCUGGGCCCCCAGGGCCUCCUGGCUUUGGGA

AACCAGGCUUGCCAGGCCUGAAGGGAGAAAGAGGACCUGCUGGCCUUCCUGGG

GGUCCAGGUGCCAAAGGGGAACAAGGGCCAGCAGGUCUUCCUGGGAAGCCAGG

UCUGACUGGACCCCCUGGGAAUAUGGGACCCCAAGGACCAAAAGGCAUCCCGG

GUAGCCAUGGUCUCCCAGGCCCUAAAGGUGAGACAGGGCCAGCUGGGCCUGCA

GGAUACCCUGGGGCUAAGGGUGAAAGGGGUUCCCCUGGGUCAGAUGGAAAAC

CAGGGUACCCAGGAAAACCAGGUCUCGAUGGUCCUAAGGGUAACCCAGGGUUA

CCAGGUCCAAAAGGUGAUCCUGGAGUUGGAGGACCUCCUGGUCUCCCAGGCCC

U GU GGGC CC AGC AGGAGC A A AGGG A AU GCC CGGAC AC A AU GGAGAGGCU GGC C

CAAGAGGUGCCCCUGGAAUACCAGGUACUAGAGGCCCUAUUGGGCCACCAGGC

AUUCCAGGAUUCCCUGGGUCUAAAGGGGAUCCAGGAAGUCCCGGUCCUCCUGG

CCCAGCUGGCAUAGCAACUAAGGGCCUCAAUGGACCCACCGGGCCACCAGGGC

CUCCAGGUCCAAGAGGCCACUCUGGAGAGCCUGGUCUUCCAGGGCCCCCUGGG CCUCCAGGCCCACCAGGUCAAGCAGUCAUGCCUGAGGGUUUUAUAAAGGCAGG

CCAAAGGCCCAGUCUUUCUGGGACCCCUCUUGUUAGUGCCAACCAGGGGGUAA

CAGGAAUGCCUGUGUCUGCUUUUACUGUUAUUCUCUCCAAAGCUUACCCAGCA

AUAGGAACUCCCAUACCAUUUGAUAAAAUUUUGUAUAACAGGCAACAGCAUU

AUGACCCAAGGACUGGAAUCUUUACUUGUCAGAUACCAGGAAUAUACUAUUU

UUCAUACCACGUGCAUGUGAAAGGGACUCAUGUUUGGGUAGGCCUGUAUAAG

AAUGGCACCCCUGUAAUGUACACCUAUGAUGAAUACACCAAAGGCUACCUGGA

U C AGGCUU C AGGGAGU GC C AUC AUCGAU CU C AC AGA A A AU GACC AGGU GU GGC

UCCAGCUUCCCAAUGCCGAGUCAAAUGGCCUAUACUCCUCUGAGUAUGUCCAC

UCCUCUUUCUCAGGAUUCCUAGUGGCUCCAAUGUGAGUACACACAGAGCUAAU

CUAAAUCUUGUGCUAGAAAAAGCAUUCUCUAACUCUACCCCACCCUACAAAAU

GC AU AU GG AGGU AGGCU G A A A AG A AU GU A AUUUUU AUUUU CU G A A AU AC AG A

UUUGAGCUAUCAGACCAACAAACCUUCCCCCUGAAAAGUGAGCAGCAACGUAA

AAACGUAUGUGAAGCCUCUCUUGAAUUUCUAGUUAGCAAUCUUAAGGCUCUU

UAAGGUUUUCUCCAAUAUUAAAAAAUAUCACCAAAGAAGUCCUGCUAUGUUA

AAAACAAACAACAAAAAACAAACAACAAAAAAAAAAUUAAAAAAAAAAACAG

AAAUAGAGCUCUAAGUUAUGUGAAAUUUGAUUUGAGAAACUCGGCAUUUCCU

UUUUAAAAAAGCCUGUUUCUAACUAUGAAUAUGAGAACUUCUAGGAAACAUC

CAGGAGGUAUCAUAUAACUUUGUAGAACUUAAAUACUUGAAUAUUCAAAUUU

AAAAGACACUGUAUCCCCUAAAAUAUUUCUGAUGGUGCACUACUCUGAGGCCU

GUAUGGCCCCUUUCAUCAAUAUCUAUUCAAAUAUACAGGUGCAUAUAUACUU

GUUAAAGCUCUUAUAUAAAAAAGCCCCAAAAUAUUGAAGUUCAUCUGAAAUG

C A AGGU GCUUUC AU C A AU GA ACCUUUUC A A ACUUUU CU AU GAUU GC AGAGA A

GCUUUUUAUAUACCCAGCAUAACUUGGAAACAGGUAUCUGACCUAUUCUUAU

UUAGUUAACACAAGUGUGAUUAAUUUGAUUUCUUUAAUUCCUUAUUGAAUCU

UAUGUGAUAUGAUUUUCUGGAUUUACAGAACAUUAGCACAUGUACCUUGUGC

CUCCCAUUCAAGUGAAGUUAUAAUUUACACUGAGGGUUUCAAAAUUCGACUA

GAAGUGGAGAUAUAUUAUUUAUUUAUGCACUGUACUGU AUUUUU AUAUUGCU

GUUUAAAACUUUUAAGCUGUGCCUCACUUAUUAAAGCACAAAAUGUUUUACC

UACUCCUUAUUUACGACGCAAUAAAAUAACAUCAAUAGAUUUUUAGGCUGAA

UU A AUUU GA A AGC AGC A AUUU GCU GUU CU C A AC C AUU CUUUC A AGGCUUUU C

AUUGUUCAAAGUUAAUAAAAAAGUAGGACAAUAAAGUGAAAAAAAAAAAAAA

AAAA (SEQ ID NO:359)

XVII. Computer Systems [00114] While various embodiments of the invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed.

Computer systems

[00115] The present disclosure provides computer systems that are programmed to implement methods of the disclosure. FIG. 13 shows a computer system 1301 that is programmed or otherwise configured to identify biomarkers for a cancer, such as a breast cancer. The computer system 1301 can regulate various aspects of the analysis of the present disclosure, such as, for example, it can analyze a cohort of biomarkers from a population of subjects afflicted with a cancer; it can identify a first subset from said cohort of said biomarkers that has at least a 3 -fold higher expression level in said cancer as compared to tissue samples that do not contain cancer, such a healthy control biomarker; it can identify a second subset from said first subset of said biomarkers that have a false discovery rate of less than a 10 ⁶ it can use at least one biomarker from said second subset of said biomarkers as input for a machine learning algorithm such as correlation feature selection (CFS); and it can further output one or more biomarkers that identify said cancer. The computer system 1301 can be an electronic device of a user or a computer system that is remotely located with respect to the electronic device. The electronic device can be a mobile electronic device.

[00116] The computer system 1301 includes a central processing unit (CPU, also “processor” and“computer processor” herein) 1305, which can be a single core or multi core processor, or a plurality of processors for parallel processing. The computer system 1301 also includes memory or memory location 1310 (e.g., random-access memory, read-only memory, flash memory), electronic storage unit 1315 (e.g., hard disk), communication interface 1320 (e.g., network adapter) for communicating with one or more other systems, and peripheral devices 1325, such as cache, other memory, data storage and/or electronic display adapters. The memory 1310, storage unit 1315, interface 1320 and peripheral devices 1325 are in communication with the CPU 1305 through a communication bus (solid lines), such as a motherboard. The storage unit 1315 can be a data storage unit (or data repository) for storing data. The computer system 1301 can be operatively coupled to a computer network (“network”) 1330 with the aid of the communication interface 1320. The network 1330 can be the Internet, an internet and/or extranet, or an intranet and/or extranet that is in communication with the Internet. The network 1330 in some cases is a telecommunication and/or data network. The network 1330 can include one or more computer servers, which can enable distributed computing, such as cloud computing. The network 1330, in some cases with the aid of the computer system 1301, can implement a peer-to-peer network, which may enable devices coupled to the computer system 1301 to behave as a client or a server. The system can train a number of classifiers that identify breast cancer.

[00117] The CPU 1305 can execute a sequence of machine-readable instructions, which can be embodied in a program or software. The instructions may be stored in a memory location, such as the memory 1310. The instructions can be directed to the CPU 1305, which can subsequently program or otherwise configure the CPU 1305 to implement methods of the present disclosure. Examples of operations performed by the CPU 1305 can include fetch, decode, execute, and writeback.

[00118] The CPU 1305 can be part of a circuit, such as an integrated circuit. One or more other components of the system 1301 can be included in the circuit. In some cases, the circuit is an application specific integrated circuit (ASIC).

[00119] The storage unit 1315 can store files, such as drivers, libraries and saved programs. The storage unit 1315 can store user data, e.g., user preferences and user programs. The computer system 1301 in some cases can include one or more additional data storage units that are external to the computer system 1301, such as located on a remote server that is in communication with the computer system 1301 through an intranet or the Internet.

[00120] The computer system 1301 can communicate with one or more remote computer systems through the network 1330. For instance, the computer system 1301 can

communicate with a remote computer system of a user (e.g., it can access electronic data from the TCGA project). Examples of remote computer systems include personal computers (e.g, portable PC), slate or tablet PC’s (e.g., Apple® iPad, Samsung® Galaxy Tab), telephones, Smart phones (e.g, Apple® iPhone, Android-enabled device, Blackberry®), or personal digital assistants. The user can access the computer system 1301 via the network 1330.

[00121] Methods as described herein can be implemented by way of machine (e.g., computer processor) executable code stored on an electronic storage location of the computer system 1301, such as, for example, on the memory 1310 or electronic storage unit 1315. The machine executable or machine readable code can be provided in the form of software.

During use, the code can be executed by the processor 1305. In some cases, the code can be retrieved from the storage unit 1315 and stored on the memory 1310 for ready access by the processor 1305. In some situations, the electronic storage unit 1315 can be precluded, and machine-executable instructions are stored on memory 1310.

[00122] The code can be pre-compiled and configured for use with a machine having a processer adapted to execute the code, or can be compiled during runtime. The code can be supplied in a programming language that can be selected to enable the code to execute in a pre-compiled or as-compiled fashion.

[00123] Aspects of the systems and methods provided herein, such as the computer system 1301, can be embodied in programming. Various aspects of the technology may be thought of as“products” or“articles of manufacture” typically in the form of machine (or processor) executable code and/or associated data that is carried on or embodied in a type of machine readable medium. Machine-executable code can be stored on an electronic storage unit, such as memory (e.g., read-only memory, random-access memory, flash memory) or a hard disk. “Storage” type media can include any or all of the tangible memory of the computers, processors or the like, or associated modules thereof, such as various semiconductor memories, tape drives, disk drives and the like, which may provide non-transitory storage at any time for the software programming. All or portions of the software may at times be communicated through the Internet or various other telecommunication networks. Such communications, for example, may enable loading of the software from one computer or processor into another, for example, from a management server or host computer into the computer platform of an application server. Thus, another type of media that may bear the software elements includes optical, electrical and electromagnetic waves, such as used across physical interfaces between local devices, through wired and optical landline networks and over various air-links. The physical elements that carry such waves, such as wired or wireless links, optical links or the like, also may be considered as media bearing the software. As used herein, unless restricted to non-transitory, tangible“storage” media, terms such as computer or machine“readable medium” refer to any medium that participates in providing instructions to a processor for execution.

[00124] Hence, a machine-readable medium, such as computer-executable code, may take many forms, including but not limited to, a tangible storage medium, a carrier wave medium or physical transmission medium. Non-volatile storage media include, for example, optical or magnetic disks, such as any of the storage devices in any computer(s) or the like, such as may be used to implement the databases, etc. shown in the drawings. Volatile storage media include dynamic memory, such as main memory of such a computer platform. Tangible transmission media include coaxial cables; copper wire and fiber optics, including the wires that comprise a bus within a computer system. Carrier-wave transmission media may take the form of electric or electromagnetic signals, or acoustic or light waves such as those generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media therefore include for example: a floppy disk, a flexible disk, hard disk, magnetic tape, any other magnetic medium, a CD-ROM, DVD or DVD- ROM, any other optical medium, punch cards paper tape, any other physical storage medium with patterns of holes, a RAM, a ROM, a PROM and EPROM, a FLASH-EPROM, any other memory chip or cartridge, a carrier wave transporting data or instructions, cables or links transporting such a carrier wave, or any other medium from which a computer may read programming code and/or data. Many of these forms of computer readable media may be involved in carrying one or more sequences of one or more instructions to a processor for execution.

[00125] The computer system 1301 can include or be in communication with an electronic display 1335 that comprises a user interface (EΊ) 1340 for providing, for example, an output listing one or more biomarkers that identify a cancer, such as breast cancer.

Examples of Ed’s include, without limitation, a graphical user interface (GUI) and web-based user interface.

[00126] Methods and systems of the present disclosure can be implemented by way of one or more algorithms. An algorithm can be implemented by way of software upon execution by the central processing unit 1305.

XIII. Kits

[00127] In one aspect, the invention provides kits comprising any of the primers and reagents for detecting the 3-gene panel of biomarkers described in this application. The kits may comprise at least one primer sequence that has at least 90% identity to any one of SEQ ID NO: 1- SEQ ID NO: 356, and a buffer solution/system. In some embodiments, the kit comprises at least one forward primer that has at least 90% identity to any one of SEQ ID NO: 1-40, SEQ ID NO:56-l50, or SEQ ID NO:228-248 and at least one reverse primer that has at least 90% identity to any one of SEQ ID NO:4l-85, SEQ ID NO: 151-227, or SEQ ID NO:249-279. In some embodiments, the kit comprises at least one forward reference primer that has at least 90% identity to any one of SEQ ID NO:280-293 or SEQ ID NO:311-324 and at least one reverse reference primer that has at least 90% identity to any one of SEQ ID NO: 294-310 or SEQ ID NO: 325-338. In some embodiments, the kit comprises at least one forward positive control primer that has at least 90% identity to any one of SEQ ID NO:339- 347 and at least one reverse positive control primer that has at least 90% identity to any one of SEQ ID NO: 348-356. In some embodiments, the kit comprises at least one forward and reverse primer sequence for each of IBSP, MMP11, and COL10A that has at least 90% identity to any of the primer combinations in Table 13, Table 14, and Table 15. In some embodiments, the kit comprises at least one forward and reverse primer sequence for each of TTC5 and C2orf44 that has at least 90% identity to any of the primer combinations in Table 16 and Table 17. In some embodiments, the kit comprises at least one forward and reverse primer sequence for chr3 gDNA that has at least 90% identity to any of the primer

combinations in Table 18.

[00128] In some instances, the kits further comprise a DNA-intercalating dye or a fluorescent probe, such as a TaqMan compatible probe. A TaqMan compatible probe may comprise a short oligonucleotide sequence designed to hybridize to the desired gene, in combination with a 5’-fluorophore and a 3’ -quencher attached to either end of the

oligonucleotide. In some instances the kit also comprises a negative control sample, a positive control sample, or a using a synthetic nucleotide control.

[00129] The kits can further comprise a set of reagents for a polymerase chain reaction. Such reagents for a polymerase chain reaction include a suitable thermostable DNA polymerase (e.g. Taq polymerase, which may be a hot-start polymerase to improve fidelity) solution, a solution of 4 dNTPs (e.g. dATP, dTTP, dGTP, dCTP), a buffer solution, DNAse- free water, and/or solutions of PCR stabilizers/enhancers. Buffers are prepared at the pH optimum for the enzyme and may additionally comprise salts such as KC1, NaCl, and/or MgCl2, reducing agents such as DTT or B-me, detergents such as triton-x or tween-20, and/or glycerol as useful for function of the enzyme. Stabilizers/additives may include agents such as DMSO, betaine monohydrate, formamide, MgCl2, glycerol, BSA, tween-20,

Tetramethyl ammonium chloride, and/or 7-deaza-2'-deoxyguanosine. For qPCR

applications, a polymerase chain reaction kit may include suitable fluorescent DNA-binding dyes such as SYBR Green, ethidium bromide, or EvaGreen.

[00130] In some examples, the set of reagents can be for a reverse-transcriptase polymerase chain reaction. Reagents for a reverse-transcriptase polymerase chain reaction include a suitable reverse transcriptase (such as Maloney murine leukemia virus, M-MLV, reverse transcriptase) solution, solution of 4 dNTPs (e.g. dATP, dTTP, dGTP, dCTP), a buffer solution, an RNAse inhibitor solution, and/or RNAse-free water. In the case of solutions for reverse-transcriptase polymerase chain reaction, all the reagents (e.g. dNTPs, water, buffer) are certified RNAse free to prevent template degradation. [00131] The kit can further comprise written instructions for a use thereof. Such instructions may include instructions for isolating/preparing the sample, operating

instrumentation (e.g. qPCR instrumentation), and/or data interpretation

[00132] The kit can further comprise components for touch-prep. Such components include poly-D-lysine coated glass slides, an RNA isolation kit, and/or spin columns (suitable for isolation/purification of RNA) and collection tubes. A minimal RNA isolation kit may comprise a solution RNAse-free sample disruption buffer, solutions of RNA isolation reagents (e.g. Trizol or phenol/chloroform or phenol/chloroform/isoamyl alcohol mixtures), RNAse-free DNase, and/or a solution of an RNAse inhibitor.

[00133] The kit can further comprise components for tumor/tissue dissociation. Such components include a) solutions of enzymes for extracellular matrix (ECM) or other protein degradation such as collagenase, trypsin, elastase, hyaluronidase, and/or papain; b) solutions for lysis of red blood cells from tissue (e.g. a hypotonic lysis buffer); c) a tissue dissociator (e.g. Miltenyi gentleMACS Octo Tissue Dissociator); d) a stabilization buffer (e.g. containing protease, DNAse, and/or RNAse inhibitors); e) a lysis buffer (a buffered solution, optionally hypotonic, containing ionic or nonionic detergents such as Triton X-100, tween-20, beta-octyl glucoside, and/or SDS).

[00134] In some embodiments, the kit is a kit for the detection of positive surgical margins. Such a kit includes components such as instructions, primers or primer

combinations outlined above (e.g. forward and reverse primers for each target gene, forward and reverse primers for a reference gene, and forward and reverse primers for a gDNA control gene), touch prep components as described above, reagents for polymerase chain reaction, and reagents for reverse-polymerase chain reaction. In some embodiments such a kit consists of as instructions, touch prep components as described above, reagents for polymerase chain reaction, and reagents for reverse-polymerase chain reaction.

[00135] In some embodiments, the kit is a kit for detection of molecular complete response (mCR). Such a kit includes components such as instructions, primers or primer combinations outlined above (e.g. forward and reverse primers for each target gene, forward and reverse primers for a reference gene, and forward and reverse primers for a gDNA control gene), components for tumor/tissue dissociation as described above, an RNA isolation kit, and/or spin columns suitable for isolation/purification of RNA.

[00136] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. It is not intended that the invention be limited by the specific examples provided within the specification. While the invention has been described with reference to the aforementioned specification, the descriptions and illustrations of the embodiments herein are not meant to be construed in a limiting sense. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. Furthermore, it shall be understood that all aspects of the invention are not limited to the specific depictions, configurations or relative proportions set forth herein which depend upon a variety of conditions and variables. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is therefore contemplated that the invention shall also cover any such alternatives, modifications, variations or equivalents. It is intended that the following claims define the scope of the invention and that methods and structures within the scope of these claims and their equivalents be covered thereby.

EXAMPLES

[00137] Example 1. A 3-gene test for residual disease/pathologic complete response

[00138] General Methods

[00139] Data from the TCGA project was analyzed to inform the identification of a cohort of biomarkers from a population of subjects afflicted with a cancer. In addition,

lumpectomies were performed on small, early-stage tumors. cDNA from these samples was prepared from clinical samples and q-PCR performed according to standard protocols. A variety of standard protocols and kits for cDNA preparation/q-PCR are known to those of skill in the art. Exemplary protocols include, those from ThermoFisher (e.g. Manual for Power SYBR® Green RNA-to-Cx^Tm l-Step Kit Part Number 4391003 Rev. D; Manual for EXPRESS One-Step; Superscript® qRT-PCR Kits, Rev. Date: 28 June 2010 Manual part no. A10327), BioRad (e.g. Manual for iTaq ¹ "¹ Universal SYBR® Green One-Step Kit 10032048 Rev B), NEB (e.g. Luna® Universal One-Step RT-qPCR Kit Protocol (E3005)), Qiagen (e.g. QuantiFast SYBR Green RT-PCR Handbook ver 07/2011, US 5,994,056, and US 6,171,785), Roche (e.g. Transcriptor One-Step RT-PCR Kit; FastStart Universal SYBR Green Master (Rox), each of which are specifically incorporated by reference herein.

[00140] Inclusion/Exclusion Criteria.

[00141] Since genomic signatures may evolve during metastasis, AJCC TNM staging (<T3) was used to restrict samples to female patients who would be eligible for a lumpectomy (see TABLE 19). One sample was excluded for failing quality control, and 8 samples were excluded for having clinical data. Inclusion/exclusion criteria preserved the racial and ethnic representation of the U.S. population, except for the only available American Indian/ Alaskan Native (AI/AN) subject, who did not satisfy our inclusion criteria. 1,014 early-stage tumors were divided (T1-T2) into a training set with 939 samples and an independent test with 75 samples. The early-stage tumor sets (Cross-validation Set and Independent Set #1) also included healthy samples from patient with late-stage tumors. This test was designed to detect early-stage tumors, but the analysis also included 175 late-stage tumors (T3-T4) as a second independent test set. In sum, novel biomarkers for cancer were identified when a computer system was used to analyze a cohort of biomarkers from the aforementioned population of subjects afflicted with a cancer. The method identified a first subset of biomarkers that had at least a 3 -fold higher expression level in said cancer as compared to a healthy control biomarker; and a second subset from said first subset of said biomarkers that provided a false discovery rate for said cancer that was less than 0.000001. The markers identified were used to train a machine learning algorithm and were experimentally validated.

[00142] The method identified a 3-gene set of markers from a plurality of biomarkers from 939 RNA Seq samples. The method was tested on two independent RNA Seq test sets (TABLE 19A and TABLE 19B). The selected 3-gene set of markers correctly classified 96.2% of 939 samples in the Cross-validation Set (early stage, AJCC 1NN4 Tumor Stages T1-12). Since these results were unexpected, we tested whether the performance estimates from cross-validation were inflated by potential modeling errors (e.g. overfitting). First, a suite of negative controls did not detect any modeling errors in the cross validation. Second, the classifier was trained on all 939 samples in the cross-validation set, and tested on a hold-out set of 75 samples. The 3-gene Random Forest test correctly classified 97.3% of 75 early-stage samples in one independent test set, and correctly classified 94.3% of 175 late-stage samples in a second independent test set. Performance was not significantly affected by race, ethnicity, tumor stage, or ER/PR/Her2 status.

By definition, overfit models have higher performance from resampling estimates like cross validation than on independent validation sets. In this case, cross validation estimates and performance on the independent validation set were within the 95% confidence intervals. These results therefore firmly exclude overfitting.

Not Available (NA), Not Evaluated (NE)

[00143] Biomarker Discovery.

[00144] A subset of biomarkers that had a large mean difference between groups, with two clearly separated distributions, was first identified using a computer system. In addition, to detect tumor cells in a population of healthy cells, additional biomarkers that had a higher level of expression in tumors than healthy samples were selected. To identify such biomarkers, genes with a log2(fold-change) = ⁺3 and genes with a False Discovery Rate (adjusted p-value) of p < 10 ⁶ were identified. The method identified a first subset of biomarkers that were overexpressed in tumors (FIG.2A). Subsequently, Correlation-based Feature Selection (CFS) was applied in the first subset from the broader cohort of biomarkers to identify genes with at least a 3 -fold higher expression level in the selected cancer as compared to a healthy control biomarker to select the top genes that contributed the most unique information.

[00145] In addition, expression of many disease-associated genes is highly correlated. Thus, the identification of genes that contribute the most unique information informed the selection of relevant markers. For instance, estrogen signaling is the classic example in breast tumors, where multiple ER-responsive genes can be the strongest biomarkers in a given tumor. But these genes would only help identify ER+ tumors, and would miss every tumor that is not ER+. Selection of highly expressed genes with CFS, or another suitable method, allows the identification of a subset of genes that not only contribute information, but that contribute the most unique information relative to the other selected genes. ETsing CFS, we selected panels of 200, 100, 20, 10, 5, 4, 3, 2, and 1 gene panels. Six machine learning algorithms were tested and performed similarly in identifying gene panels. The 6 algorithms were the support vector algorithm SMO, Naive Bayes, J48 Decision Tree, Lazy-IBk, the Multilayer Perceptron neural network, and Random Forest. The Random Forest ensemble machine learning method was used in the remaining of the experiments. There are at least 9 published classifiers for breast cancer that use gene expression, including OncoTypeDX® and PAM50. Principal Component Analysis (PCA) (FIG.3) suggests a rationale for why the disclosed 3-gene set of markers had higher performance than existing breast cancer disease classifiers. Existing classifiers attempt to identify subgroups among the cluster of tumor samples. This leads to the strongest performance being focused on distinguishing the two most prominent groups, such as tumor and healthy (FIG. 3).

[00146] The 3 -gene test had an accuracy of 94-97% when analyzed on 3 sets of RNA

Seq. samples (TABLE 20). PAM50 could not be used for margin analysis because it requires samples with >50% tumor.

[00147] TABLE 20: Comparison of 3-gene test with PAM50 on RNA Seq samples

[00148] Example 2. Test performance and validation

[00149] 1 ,014 RNA Seq samples were divided from early stage breast tumors and adjacent healthy tissue into a Cross-validation set of 939 samples and an Independent Test Set of 75 samples (TABLE 19). The 3-gene test correctly classified 96.2% of the samples in the Cross-validation set (TABLE 20). The Area Linder the Receiver Operator Characteristic Curve (ALiC ROC) was 0.990 (95% Cl: 0.997-1.000) (FIG.4). The 3-gene test has equivalent performance on the early-stage Independent Test Set: 97.3% Accuracy, 0.998 AUC ROC (95% Cl: 0.992-1.000), 98.0% Sensitivity, 96.0% Specificity, 98.0% Positive Predictive Value, and 96.0% Negative Predictive Value.

[00150] To validate the test in other tumor samples, T3 and T4 (later stage) samples, were also tested. 175 late-stage primary tumors were used as a second independent test set. In this analysis, the classifier correctly detects 94.3% of late-stage tumors. In all tests sets, the 3- gene test performed equally well regardless of racial groups or clinical subtypes (ER±, PR, Her2±) (see TABLE 19).

[00151] 10-Fold Cross validation (CV).

[00152] Our classifier combines a 3 -gene set of markers including MM PI /, IBSP, and

COL10AX , using the Random Forest machine learning algorithm. We used ten-fold CV to estimate performance with RNA Seq data. For ten-fold CV, each of the 939 samples was used once (and only once) in 1 of 10 independent validation sets. The first iteration used subsets 2-10 (S2.10) as a training set (Tl), while subset 1 (Si) was withheld as the validation set (VI). This was repeated on a total of 10 independent validation sets.

[00153] Negative Controls for Cross Validation.

[00154] After analyzing test performance, a panel of four separate negative controls was created to demonstrate that the modelling was performed correctly.

[00155] Negative Control I: Randomized. Fictitious Class Labels to Detect Overfitting

[00156] Biomarker selection workflow was performed on a dataset with randomized class labels to detect overfitting. Using the existing classification of samples in the dataset as either Tumor or Healthy, markers were randomly assigned to a fictitious class or to a gene expression class (Class A or B). The workflow was repeated in the same manner used to develop the 3-gene test, this time trying to distinguish Class A from B. The 3 best genes were selected in each of 10 cross validation folds, and used Random Forest to train a classifier for each fold. Subsequently, 10 independent test sets were used to determine performance of the 10 models. By performing the disclosed workflow on a dataset with randomized class labels, the strategy detects overfitting. FIG.6, Negative Control I (Random Class) clearly shows no evidence of overfitting: 0.51 Area Under the ROC Curve, 51.9% Accuracy, 51.6%

Sensitivity, 52.1% Specificity.

[00157] Negative Control II: Randomly Selected Genes

[00158] 3 randomly selected genes were modeled in each of 10 cross validation folds

(FIG.6): 0.733 AUC ROC, 72.6% Accuracy, 73.8% Sensitivity, 61.8% Specificity, however randomly selected genes perform much worse than our 3-Gene Test. This data demonstrates that randomly selected genes do not provide an adequate set of biomarkers.

[00159] Negative Control III: Reversed Selection Criteria

[00160] Poorly performing genes were selected by reversing the disclosed selection criteria: i.e., genes with poor p-values and small differences between tumor and healthy samples were selected in the reversed selection criteria (FIG.6): 0.653 AUC ROC, 61.9% Accuracy, 61.5% Sensitivity, and 65.4% Specificity. The ROC plot confirms that poor genes provide less information than other disclosed genes.

[00161] Negative Control IV: Comparison to Null Model

[00162] We applied a Null Model to our dataset and compared the performance of our 3-gene panel to it. The Null Model consistently guesses that each sample is a member of the most prevalent class; in this case, that each sample is tumor (FIG.6, dashed line). The 3-gene test (FIG.6, dark line) has a p-value smaller than 5 x 10 ¹¹ compared to the Null Model. While this provides a performance benchmark, it also models the diagnostic performance of a treatment strategy. Some well-intentioned surgeons proposed taking additional tissues (routine second margins) from all patients. Like routine second margins, the Null Model assumes that all patients have positive margins. FIG.4 shows how the sensitivity and specificity of the Null Model (dashed line) compares to microscopy.

[00163] Example 3. Assay development for qPCR test

[00164] Primer Design.

[00165] Hundreds of primers were designed for use in in-silico evaluations.

Approximately 40 primer pairs were synthesized and tested empirically using synthetic cDNA template. For all experiments, we used clinical-grade reagents and pipettes that are certified to 1S08655 standards. qPCR reagents were manufactured in cGMP conditions under ISO9001 management in a facility that is IS0l3485-registered.

[00166] The disclosed one-step RTqPCR assay uses targeted primers to reverse transcribe RNA into cDNA, followed by qPCR amplification of cDNA and detection using a DNA-intercalating dye. Synthetic templates were utilized to optimize the concentration of each primer (titrations of primer concentrations) and annealing temperatures (temperature gradients). Some RNA primers were designed to span exon junctions. For exon-spanning primers, genomic DNA from HeLa cells was used to verify that RNA quantification is not impacted by the presence of genomic DNA. For each primer pair, a synthetic template was used to determine performance parameters: lO-fold dilutions (5 technical replicates of 6 concentrations), 2-fold dilutions (7 technical replicates of 6 concentrations), and 24 no- template controls. Lastly, pooled RNA from 3 invasive breast tumors was used to test each primer pair. The testing evaluated 3 tumor-specific genes ( IBSP , MMP1J and COL10A1 ), 2 reference genes ( C2orf44 and TTC5 ), and a control to detect genomic DNA, chr3 gDNA. For each primer pair, 2 negative controls were included and 3 technical replicates were conducted. All RNA experiments also included a positive control for each primer pair.

Altogether, assay development and validation involved greater than 3,700 reactions of 20- microliter reactions.

[00167] Absolute Quantification of 22 Clinical Samples.

[00168] The disclosed qPCR assays were used to analyze RNA from 22 clinical samples (11 pairs of invasive breast adenocarcinomas and adjacent healthy samples).

Specificity, prevision, sensitivity, linearity, and PCR efficiency were determined for the top qPCR primer combinations, which were all designed to span exon junctions (TABLES 1-6). Performance criteria were found to satisfy MIQE and CLSI guidelines.

[00169] TABLE 21: Analytical Validation of qPCR primers

[00170] FIGURE 7A and FIGURE 7B depict charts showing analytic validation of qPCR assays for using clinical-grade reagents. FIGURE 7A panel a depicts amplification plots of 20 microliter qPCR reactions. 12 concentrations of synthetic cDNA template (1.1 million to 0 copies per microliter), including lO-fold dilutions for 6 high concentrations (5 technical replicates) and 2-fold dilutions for 5 low concentrations (7 technical replicates).

One concentration point overlapped in the high and low concentration series. Each primer pair includes 24 replicates of no-template controls. Error bars at each cycle represent 95% Cl of technical replicates.

[00171] FIGURE 7A panel b depict fluorescence versus cycle plots to determine Ct iorMMPll. A 4-parameter linear model was fitted to 5 technical replicates (circles). The maximum of the second derivative was used to define the Ct (CtD2).

[00172] FIGURE 7B panel c depicts threshold cycle versus template dilution plots to calculate linear range. The linear range is defined as the range of concentrations where CtD2 fit a straight line with R-squared >0.995. Red lines indicate 95% Confidence Intervals calculated from 200 bootstraps. FIGURE 7B panel d depicts melt plots confirm to specificity of the primers. Increasing temperature denatures PCR amplicons, which decreases fluorescence. A single peak of the negative first derivative confirms the presence of a single amplicon. The peak corresponds to the expected melting temperature (dashed line).

[00173] FIGURE 7A and FIGURE 7B panels e-h depict charts showing analytic validation of qPCR assays for IBSP RNA as for A/A//^J / / All assays used clinical-grade reagents. Panel e depicts amplification plots of 20 microliter qPCR reactions. 12

concentrations of synthetic cDNA template (1.1M to 0 copies per microliter), including 10- fold dilutions for 6 high concentrations (5 technical replicates) and 2-fold dilutions for 5 low concentrations (7 technical replicates). One concentration point overlapped in the high and low concentration series. Each primer pair includes 24 replicates of no-template controls. Error bars at each cycle represent 95% Confidence Intervals of technical replicates.

[00174] FIGURE 7A Panel f depicts fluorescence versus cycle plots to determine Ct for IBSP. A 4-parameter linear model was fitted to 5 technical replicates (circles). The maximum of the second derivative was used to define the Ct (CtD2). FIGURE 7B panel g depicts threshold cycle versus template dilution plots to calculate linear range. The linear range is defined as the range of concentrations where CtD2 fit a straight line with R-squared >0.995. Red lines indicate 95% Confidence Intervals calculated from 200 bootstraps.

[00175] Absolute quantification (qPCR) of RNA from 22 clinical samples confirms that biomarker expression is substantially higher in tumors (TABLE 21). We present adjusted copy numbers for all tumor results from this experiment, including TABLE 21 and FIG.8. If a tumor sample did not contain 100% tumor (e.g. 95% tumor), the estimated number of copies was adjusted to the equivalent of 100% tumor. The mean number of RNA copies was 34 to 176 times higher in tumor than healthy. On average, expression was 72 to 189 times higher when we compared each tumor to paired healthy tissue from the same patient. One advantage of multi-analyte modeling is that all genes do not need to be elevated in every patient (otherwise the test would only require 1 gene). In each pair of samples, at least one of the tumor biomarkers was markedly elevated. On average, the best of the 3 genes was 273 times higher in the tumor sample than the paired healthy sample.

[00176] TABLE 22. Absolute quantification (RTqPCR) of RNA from 22 tumors and healthy tissues.

[00177] qPCR Test Performance

A panel of 3 biomarkers correctly classified 100% of the samples as tumor or healthy using Random Forest (EXAMPLE 4), Generalized Linear Models of the Binomial Family

(EXAMPLE 5), and Regularized Discriminant Analysis (EXAMPLE 6). FIG.9 shows the ROC curve for a 3-gene test using Random Forest. In addition, IBSP RNA and MMP11 RNA can unexpectedly be used in combination in a Generalized Linear Model of the Binomial Family to correctly classify 100% of samples (EXAMPLE 7). We further demonstrated that IBSP RNA (FIG.10) and MMP11 RNA (FIG.11), correctly classified 100% of samples when used individually in a Generalized Linear Model of the Binomial Family (EXAMPLE 7). Additionally, when COL10A1 RNA was used as an individual biomarker in a Generalized Linear Model of the Binomial Family, the disclosed qPCR assay correctly classified 77.3% of samples as tumor of healthy.

[00178] Example 4. Performance of the 3-gene test using the Random Forest ( R G ) machine learning algorithm, as determined by 5-fold cross validation.

[00179] 22 samples were analyzed using the disclosed RTqPCR assay as follows:

[00180] 2 classes of samples were analyzed: RNA from 11 tumor samples and 11 healthy samples were analyzed using the disclosed clinical-grade RTqPCR assays.

Resampling was used to estimate performance and statistical parameters of a test generated using Random Forest. Five-fold cross validation showed that the 3-gene RF test had an accuracy of 100%, as shown in TABLE 24.

[00181] The following parameters were used or determined in the analysis:

[00182] Example 5. Performance of the 3-gene test using a Generalized Linear Model .

[00183] 22 samples were analyzed using the disclosed RTqPCR assay as follows:

[00184] 2 classes of samples were analyzed: RNA from 11 tumor samples and 11 healthy samples were analyzed using the disclosed clinical-grade RTqPCR assays.

Resampling was used to estimate performance and statistical parameters of a test generated using a Generalized Linear Model in the binomial family. Five-fold cross validation showed that the 3-gene glm test had an accuracy of 100%, as shown in TABLE 26.

[00185] The following describes R output for this analysis:

[00186] Example 6. Performance of the 3-gene test using a Regularized

Discriminant Analysis (RDA).

[00187] 22 samples were analyzed using the disclosed RTqPCR assay as follows:

[00188] 2 classes of samples were analyzed: RNA from 11 tumor samples and 11 healthy samples were analyzed using the disclosed clinical-grade RTqPCR assays.

Resampling was used to estimate performance and statistical parameters of a test generated using Regularized Discriminant Analysis (RDA). Five-fold cross validation showed that the 3-gene RDA test had an accuracy of 100%, as shown in TABLE 27.

[00189] The following describes R output for this analysis:

[00190] Accuracy was used to select the optimal model using the largest value. The final values used for the model were gamma = 0 and lambda = 0.

[00191] Example 7. Performance of the 2-gene test using a Generalized Linear Model (glm).

[00192] 22 samples were analyzed using the disclosed RTqPCR assay as follows:

[00193] 2 classes of samples were analyzed: RNA from 11 tumor samples and 11 healthy samples were analyzed using the disclosed clinical-grade RTqPCR assays. The two genes were IBSP and MMP I / Resampling was used to estimate performance and statistical parameters of a test generated using a Generalized Linear Model (glm). Five-fold cross validation showed that the 2-gene glm test had an accuracy of 100%, as shown in TABLE 29.

[00194] The following describes R output for this analysis:

[00195] Example 8. Analysis of Validation Data for 3-gene test

[00196] Data from 1,211 samples using 3 different technologies was used to validate a 3-gene test. Cross-validation was performed using 939 RNA Seq samples. Since cross- validation uses independent test sets, one can mathematically prove that it is a reliable estimate of performance. This can be confirmed with a suite of negative controls. Similar expression patterns have been confirmed with RNA Seq and microarray, confirming that the signals are not platform specific. Independent test sets were then analyzed to determined performance on early and late stage tumors, specifically, test sets 1 (75 early-stage samples) and independent test set 2 (175 late-stage samples).

[00197] The results of these analyses were extremely surprising in light of the current beliefs about breast cancer biology. To further investigate these surprising results, we used clinical-grade reagents to develop and validate RTqPCR assays for the selected mRNAs. We analyzed 22 samples with our assays. This represents the third independent set of samples. Using the disclosed RTqPCR assays, the selected biomarkers clearly demarcate tumor and healthy, and even provides more separation between tumor and healthy than RNA Seq.

[00198] We used predictive models to combine the biomarkers and set clinically actionable thresholds. We discovered that multiple predictive models can achieve excellent performance. For example, when all 3 biomarkers are detected using the disclosed RTqPCR assays, Random Forest, Regularized Discriminant Analysis, and Generalized Linear Models of the Binomial Family can correctly classify 100% of the samples as tumor or healthy. In addition, we unexpectedly discovered that when IBSP RNA (FIG.10) and MMP11 RNA (FIG.11) are detected using the disclosed assays, they can be used individually to correctly classify 100% of samples. These experiments firmly establish that the signal is tumor-specific and reproducible across three detection technologies.

[00199] Example 9. Evaluation of surgical margins

[00200] As demonstrated herein, the markers and methods can be used to improve the evaluation of surgical margins. In this case, cells are collected from the surface of a surgical specimen and the disclosed assays are used to detect the disclosed markers. A number of methods can be used to collect cells from the surface of a surgical specimen. As non-limiting examples, cells can be collected using a surface with a functionalized surface, such as a poly lysine coated touch imprint cytology slide. Cells could also be collected using a membrane, such as a nitrocellulose membrane. In addition, cells could be collected using a sharp or blunt instrument, such as scrape preparations, which are routinely performed for pathologic examination.

[00201] The markers and methods could be used to screen patients for invasive breast cancer. Specimens can be collected using nipple aspirates or ductal lavage, where the mammary ducts and glands are flushed with fluid and aspirated, sometimes following brief hormonal stimulation. Existing screening methods suffer from poor sensitivity or specificity, and often exposure patients to radiation. Ductal lavage is the preferred screening method for some surgeons, because it directly samples the ducts and glands that give rise to epithelial tumors like adenocarcinomas. However, the analysis of rare tumor cells is not ideal.

Microscopic detection of tumors has the best performance when the tumor is analyzed in the context of its surrounding healthy tissue. In fact, the name histo-pathology derives from the Greek histos , meaning tissue. When cells are scraped or flushed into a suspension, they are no longer in the context of surrounding tissue. Ductal lavage is therefore a promising screening strategy that is currently limited by the microscopic analysis required to detect rare or isolated breast cancer cells. Molecular analysis is particularly well suited to solve this problem because it does not rely on visual analysis, and does not require tumor to be evaluated in the context of healthy tissue. The disclosed markers and methods could therefore be used as a screening tool to determine whether there are invasive cancer cells present in screened patients.

[00202] The markers and methods could be used to detect or diagnose invasive adenocarcinoma from biopsies of the breast. Biopsies could include core biopsies, punch biopsies, incisional biopsies and excisional biopsies. In many biopsy samples, the procedure did not collect a sufficient amount of cells, or the tissue architecture has been disrupted, making it challenging to reach a definitive histopathologic or cytological diagnosis. These challenging cases are prime examples of the advantage of molecular analysis. Molecular analysis does not require abundant tissue, and does not require intact tissue structures in order to detect the disclosed signatures of invasive cancer.

[00203] Example 10. Identifying pre-cancerous lesions

[00204] The disclosed markers and methods can be used to establish a new diagnostic paradigm for pre-cancerous lesions. Lesions like ductal carcinoma in situ (DCIS) and lobular carcinoma in situ (LCIS) are currently considered pre-cancerous lesions or risk factors for invasive cancer. In only some cases do they develop into invasive cancer, but there is currently no way to identify which lesions have invasive potential. Moreover, precursor lesions are only analyzed by a few microscopic sections. The current diagnostic paradigm for precancerous lesions is based on whether a pathologist happens to observe cells that penetrate the basement membrane on the few slides that they examine. There is therefore thought to be a subset of pre-cancerous lesions with undiagnosed invasive potential. The disclosed markers and methods provide a molecular analysis of invasiveness that could identify those precancerous lesions with invasive potential. In addition, the disclosed methods can be performed on a more representative portion of the specimen than 3 microscopic sections. As non-limiting examples, tissue or biopsy specimens can be morcellated, digested

enzymatically, and/or chemically lysed to release the disclosed biomarkers, which can then be detected using the disclosed methods. In this way, the disclosed biomarkers represent a strategy to stratify patients by their risk for developing invasive cancer.

[00205] Pathologic Complete Response (pCR) is the absence of residual cancer in a solid tissue specimen, obtained from a patient who was previously diagnosed with invasive cancer. pCR is used as a surrogate endpoint for solid tumor neoadjuvant therapies. However, FDA guidance acknowledges that there is an“uncertain relationship between pCR and long term outcome,” and emphasizes the possibility“that a neoadjuvant trial could fail to demonstrate a significant difference in pCR rates and result in abandoned development of a drug that is, in fact, active in the adjuvant or metastatic setting.” A 2016 analysis found that pCR is the primary endpoint of -50% of enrolling phase II rectal cancer trials, and 45% of phase III preoperative breast cancer trials. However, there are reasons to continually improve the metrics and technologies that serve as surrogates for long-term outcomes. Hormonal therapies exemplify treatments that substantially improve survival, with only minimal impacts on pCR. Conversely, pertuzumab was approved by the FDA following a phase II randomized clinical trial that demonstrated an improvement in pCR. Yet, to date, there have been no data that suggest pertuzumab improves event-free survival, disease-free survival, or overall survival in the neoadjuvant setting. These cautionary notes underscore the importance of efforts to vigorously improve the detection of minimal residual disease and the need to develop a molecular complete response (mCR) assay.

[00206] Histopathology has been the best way to examine tumors for over a century, but it is not ideal to hunt for minimal residual disease (MRD). While FDA guidance documents emphasize the importance of compressive sectioning, sampling by pathology is woefully underpowered to provide a statistically meaningful analysis of the specimen (e.g. in practice, only a few sections are used to hunt for elusive residual tumor).

[00207] Detecting metastases to lymph nodes exemplifies the challenges of detecting breast cancer MRD using microscopy. Donald Weaver wrote that“It is quite clear that the more sections we evaluate from SLNs the more metastases we identify; however, it is impractical to expect the practicing pathologist to mount, stain, and microscopically examine every section through the SLN paraffin blocks.” Nevertheless,“when we fail to examine the entire node, we as pathologists miss metastases that are present.”

[00208] Older recommendations of sectioning lymph nodes in intervals is no longer considered appropriate because thicker intervals (3-4mm intervals) mean less metastases are detected. Examining a greater number of thinner sections detects more metastases. When thin sectioning was adopted in the United States between 1995 to 1999, node positive Stage II breast cancer increased from 60 to 80 cases per 100,000 population-based individuals in the SEER national cancer database. [00209] EMBODIMENTS

[00210] Embodiment 1. A method of distinguishing a cancer from adjacent healthy tissue, said method comprising: a) obtaining a specimen from a human subject, b) collecting a sample from said specimen, c) detecting a presence of a set of markers in said sample by performing an amplification reaction in a plurality of polynucleotides from said sample, wherein said set of markers is selected from the group consisting essentially of: Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain ( COL10A1 ); and d) distinguishing said cancer when a threshold level of said set of markers is detected.

[00211] Embodiment 2. The method of Embodiment 1, wherein said amplification reaction is a PCR reaction.

[00212] Embodiment 3. The method of Embodiment 1, wherein said PCR reaction is a qPCR reaction.

[00213] Embodiment 4. The method of Embodiment 1, wherein said PCR reaction is a RTqPCR reaction.

[00214] Embodiment 5. The method of Embodiment 1, wherein said method can distinguish said cancer in at least lOng of said plurality of polynucleotides from sample.

[00215] Embodiment 6. The method of Embodiment 1, wherein said method can distinguish said cancer in at least 100 mg of said sample.

[00216] Embodiment 7. The method of Embodiment 1, wherein said amplification reaction uses at least one primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356.

[00217] Embodiment 8. The method of Embodiment 1, wherein said sample is frozen.

[00218] Embodiment 9. The method of Embodiment 1, wherein said sample is a biopsy sample.

[00219] Embodiment 10. The method of Embodiment 9, wherein said biopsy is a liquid biopsy.

[00220] Embodiment 11. The method of Embodiment 9, wherein said biopsy is a solid tissue biopsy.

[00221] Embodiment 12. The method of Embodiment 1 [00209], wherein said cancer is breast cancer.

[00222] Embodiment 13. The method of Embodiment 12, wherein said breast cancer is invasive breast cancer. [00223] Embodiment 14. The method of Embodiment [00221], wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 96% accuracy.

[00224] Embodiment 15. The method of Embodiment [00221], wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 96% sensitivity.

[00225] Embodiment 16. The method of Embodiment [00221], wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 94% specificity.

[00226] Embodiment 17. The method of Embodiment 1, wherein said cancer is a urothelial carcinoma.

[00227] Embodiment 18. The method of Embodiment [00209], further comprising outputting a percentage of said plurality of polynucleotides expressing said markers from said sample.

[00228] Embodiment 19. The method of Embodiment [00209], further comprising comparing said set of markers from said sample to said set of markers from said a control sample.

[00229] Embodiment 20. The method of Embodiment [00227], wherein said control sample is a second sample from said human subject.

[00230] Embodiment 21. The method of Embodiment 21, further comprising performing a second assay to distinguish said cancer.

[00231] Embodiment 22. The method of Embodiment 21, wherein said second assay is an immunohistochemistry assay.

[00232] Embodiment 23. The method of Embodiment [00209], wherein said threshold level of saidA7 P77 is 1,000 copies.

[00233] Embodiment 24. The method of Embodiment 1, wherein said threshold level of said 7¾SP is 25 copies.

[00234] Embodiment 25. The method of Embodiment [00209], wherein said threshold level of said COL10A1 is 700 copies.

[00235] Embodiment 26. The method of Embodiment [00209], wherein said set of markers is selected from the group consisting of: Matrix Metallopeptidase 11 (L7MR77), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain ( COL10A1 ).

[00236] Embodiment 27. A kit comprising, at least one primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356 and a buffer system. [00237] Embodiment 28. The kit of claim [00235], wherein said buffer system is a PCR buffer system.

[00238] Embodiment 29. Isolated nucleic acid comprising a primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356.

[00239] Embodiment 30. A method of identifying a biomarker for a cancer

comprising:

(a) analyzing, by a computer system, a cohort of biomarkers from a population of subjects afflicted with a cancer;

(b) identifying, by said computer system, a first subset from said cohort of said biomarkers that has at least a 3 -fold higher expression level in said cancer as compared to a healthy control biomarker;

(c) identifying, by a computer system, a second subset from said first subset of said biomarkers that provides a false discovery rate for said cancer that is less than a 10 ⁶ rate;

(d) instructing said computer system to use at least one biomarker from said

second subset of said biomarkers as a training set of a machine learning algorithm; and

(e) outputting one or more biomarkers that identify said cancer within a 95%

confidence interval.

[00240] Embodiment 31. The method of Embodiment 30, wherein said cancer is breast cancer.

[00241] Embodiment 32. The method of Embodiment 31, wherein said breast cancer is invasive breast cancer.

[00242] Embodiment 33. The method of Embodiment 30, wherein said one or more biomarkers identify said cancer with greater than 96% accuracy.

[00243] Embodiment 34. The method of Embodiment 30, wherein said one or more biomarkers identify said cancer with greater than 96% sensitivity.

[00244] Embodiment 35. The method of Embodiment 30, wherein said one or more biomarkers identify said cancer with greater than 94% specificity.

[00245] Embodiment 36. The method of Embodiment 30, wherein said training set comprises one or more markers selected from the group consisting essentially of: Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain ( COL10A1 ). [00246] Embodiment 37. A method of diagnosing a cancer in a human subject, said method comprising:

(a) obtaining a sample from said human subject,

(b) detecting whether one or more markers are present in said sample by performing an amplification reaction in a plurality of polynucleotides from said sample and detecting the presence of said one or more markers, wherein said one or more markers are selected from the group consisting of: Matrix Metallopeptidase 11 (MMP11), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain ( COL10A1 ); and

(c) distinguishing said cancer when a threshold level of one or more said markers is detected.

[00247] Embodiment 38. A method of detecting Matrix Metallopeptidase 11 (MMP11) in a human subject, said method comprising:

(a) obtaining a sample from said human subject; and

(b) detecting whether MMP11 is present in said sample by performing an amplification reaction in a plurality of polynucleotides from said sample and detecting the presence of an MMP 11 transcript.

[00248] Embodiment 39. A method of detecting integrin binding sialoprotein (IBSP) in a human subject, said method comprising:

(c) obtaining a sample from said human subject; and

(d) detecting whether IBSP is present in said sample by performing an amplification reaction in a plurality of polynucleotides from said sample and detecting the presence of an IBSP transcript.

[00249] Embodiment 40. A method of detecting collagen type X alpha 1 chain (COLIOAI) in a human subject, said method comprising:

(e) obtaining a sample from said human subject; and

(f) detecting whether COLIOAI is present in said sample by performing an amplification reaction in a plurality of polynucleotides from said sample and detecting the presence of an COLIOAI transcript.

[00250] While preferred embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions will now occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention. It is intended that the following claims define the scope of the invention and that methods and structures within the scope of these claims and their equivalents be covered thereby.

Claims (52)

CLAIMS WHAT IS CLAIMED IS:

1. A method of distinguishing a cancer from adjacent healthy tissue, said method comprising:

(a) obtaining a specimen from a human subject,

(b) detecting a presence of a set of markers in said specimen by performing an

amplification reaction in a plurality of polynucleotides from said specimen, wherein said set of markers is selected from the group consisting essentially of: Matrix Metallopeptidase 11 ( MMP11 ), integrin binding sialoprotein ( IBSP ), and collagen type X alpha 1 chain ( COL10A1 ); and

(c) distinguishing said cancer when a threshold level of said set of markers is detected.

2. The method of claim 1, wherein said plurality of polynucleotides comprise

RNA.

3. The method of claim 1, wherein said plurality of polynucleotides comprise cDNA.

4. The method of claim 1, wherein said plurality of polynucleotides comprise

DNA.

5. The method of claim 1, wherein the detecting comprises using a DNA- intercalating dye.

6. The method of claim 1, wherein the detecting comprises using a fluorescent probe.

7. The method of claim 6, wherein said fluorescent probe is a TaqMan probe.

8. The method of claim 1, wherein said amplification reaction is a PCR reaction.

9. The method of claim 1, wherein said PCR reaction is a qPCR reaction.

10. The method of claim 1, wherein said PCR reaction is a RTqPCR reaction.

11. The method of claim 1, wherein said method can distinguish said cancer in at least lOng of said plurality of polynucleotides from specimen.

12. The method of claim 1, wherein said method can distinguish said cancer in at least 250 cells of said specimen.

13. The method of claim 1, wherein said amplification reaction uses at least one primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356.

14. The method of claim 13, wherein said amplification reaction uses at least one primer selected from SEQ ID NO: 1- SEQ ID NO: 356 to convert RNA into cDNA.

15. The method of claim 14, wherein said at least one primer pair selected from SEQ ID NO: 1- SEQ ID NO: 356 amplifies a cDNA specimen.

16. The method of claim 1, wherein said specimen is a frozen specimen, a fresh specimen, or a fixed specimen.

17. The method of claim 1, wherein said specimen is a biopsy specimen.

18. The method of claim 17, wherein said biopsy is a liquid biopsy.

19. The method of claim 17, wherein said biopsy is a solid tissue biopsy or a surgical excision.

20. The method of claim 1, wherein said specimen is obtained by imprint cytology.

21. The method of claim 20, wherein said imprint cytology is a touch-preparation.

22. The method of claim 1, wherein said specimen is obtained by scrape preparation, a nipple aspiration, or a ductal lavage.

23. The method of claim 1, wherein said cancer is breast cancer.

24. The method of claim 23, wherein said breast cancer is selected from the group consisting of: invasive adenocarcinoma, invasive ductal breast cancer, and invasive lobular breast cancer.

25. The method of claim 24, wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 90% accuracy.

26. The method of claim 24, wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 90% sensitivity.

27. The method of claim 24, wherein said method distinguishes said breast cancer from adjacent healthy tissue with greater than 90% specificity.

28. The method of claim 1, wherein said method quantitates an amount of said cancer.

29. The method of claim 28, further comprising outputting a percentage of said plurality of polynucleotides expressing said markers from said specimen.

30. The method of claim 1, further comprising comparing said set of markers from said specimen to said set of markers from said control specimen.

31. The method of claim 30, wherein said control specimen is a second specimen from said human subject or a synthetic nucleotide control.

32. The method of claim 1, further comprising performing a second assay to distinguish said cancer.

33. The method of claim 32, wherein said second assay is an

immunohistochemistry assay.

34. The method of claim 1, wherein said threshold level of said A7MP77 is 1,000 copies per microliter.

35. The method of claim 1, wherein said threshold level of said IBSP is 25 copies per microliter.

36. The method of claim 1, wherein said threshold level of said COL10A1 is 700 copies per microliter.

37. The method of claim 1, wherein said set of markers is selected from the group consisting of: Matrix Metallopeptidase 11 (MMP11), integrin binding sialoprotein (IBSP), and collagen type X alpha 1 chain (COLIOAI).

38. The method of claim 1, wherein said amplification reaction is a singleplex reaction.

39. The method of claim 1, wherein said amplification reaction is a multiplex reaction.

40. A kit comprising, at least one primer sequence that has at least 90% identity to any one of SEQ ID NO: 1- SEQ ID NO: 356, and a buffer system.

41. The kit of claim 40, wherein said buffer system is a PCR buffer system.

42. Isolated nucleic acid comprising a primer sequence that has at least 90% identity to SEQ ID NO: 1- SEQ ID NO: 356.

43. A method of identifying a biomarker for a cancer comprising:

(a) analyzing, by a computer system, a cohort of biomarkers from a population of subjects afflicted with a cancer;

(b) applying, by said computer system, a first filter to said cohort of said biomarkers to identify a first subset of biomarkers from said cohort that has at least a 3 -fold higher expression level in said cancer as compared to a healthy control biomarker;

(c) applying, by said computer system, a second filter to said first subset of biomarkers to identify a second subset of biomarkers that have a false discovery rate for said cancer that is less than 0.000001; and

(d) applying, by said computer system, a correlation based filter selection to said second subset of biomarkers to identify the biomarkers that classify the largest number of different types of said cancer.

44. The method of claim 43, wherein said correlation based filter is an anti correlation based method.

45. The method of claim 43, further comprising using the identified biomarkers as features input into a machine learning algorithm that distinguishes clinical specimens based on predefined attributes.

46. The method of claim 43, wherein said cancer is breast cancer.

47. The method of claim 46, wherein said breast cancer is selected from the group consisting of: invasive adenocarcinoma, invasive ductal breast cancer, and invasive lobular breast cancer.

48. The method of claim 43, wherein said one or more biomarkers identify said cancer with greater than 90% accuracy.

49. The method of claim 43, wherein said one or more biomarkers identify said cancer with greater than 90% sensitivity.

50. The method of claim 43, wherein said one or more biomarkers identify said cancer with greater than 90% specificity.

51. The method of claim 43, wherein said one or more biomarkers are therapeutic targets.

52. The method of claim 43, wherein said false discovery rate is a />-value for said cancer that is less than 0.0000001.