Method and system for evaluating tumor formation risk and tumor tissue source

Abstract

Provided are a tumor risk evaluation method and system. Specifically provided are a method and / or system for evaluating the correlation between a sample under test and a tumor formation risk and / or tumor tissue source. Methylation variation regions of various different cancers and specific methylation characteristic regions of various organs are captured by using DNA or RNA oligonucleotide sequences, the existence of tumor components (ctDNA) in blood cell-free DNA (cfDNA) is determined, and the correlation between the sample and the tumor tissue source is evaluated. Provided is a low-cost and high-accuracy method, which is conducive to accurately predicting and evaluating the risk of various cancers.

Description

A method and system for evaluating tumor formation risk and tumor tissue origin Technical Field

[0001] The present application relates to the field of biomedicine, and specifically to a method and system for evaluating the risk of tumor formation and the origin of tumor tissue. Background Art

[0002] DNA methylation is known to play an important role in the regulation of gene expression. Abnormal DNA methylation marks have been reported in the development and progression of many diseases, including cancer. As a high-resolution, high-throughput technology, DNA methylation sequencing is increasingly recognized for its role in cancer screening, diagnosis, and monitoring. Whole genome bisulfite sequencing (WGBS) is the gold standard for methylation sequencing, but it is difficult to use in clinical practice due to the severe damage to DNA during the processing and the high sequencing costs. More importantly, most regions of the human genome are not active during the development and progression of cancer, and cancer-related mutations are often concentrated in certain specific regions, such as CpG islands, which provides a good opportunity for targeted sequencing.

[0003] Despite this, the discovery and screening of cancer-related Differentially Methylated Regions (DMRs) is challenging because population heterogeneity, including diseases, age and other conditions, can lead to non-specific changes in the methylation spectrum. Therefore, these non-cancerous but abnormal signals need to be processed during the establishment of the cancer detection DOC (Detection of Cancer) model. Finally, for the application of detection of multiple cancer types, the establishment of the tissue tracing TOO (Tissue of Origin) model is of great auxiliary significance for tracing the possible source organs of cancer mutations, determining downstream diagnosis and treatment pathways, and saving medical costs.

[0004] Summary of the Invention

[0005] This application establishes a low-cost, high-precision method that uses DNA or RNA oligonucleotide sequences to capture methylation variation regions of various cancers and specific methylation characteristic regions of various organs, and to determine the presence of tumor components (ctDNA) in blood-free DNA (cfDNA), and to evaluate the correlation between the sample and the tumor tissue source.

[0006] On the one hand, the present application provides a method for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the origin of tumor tissue, which comprises: (1) a differential methylation region DMR division step: based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites, determining multiple target DMRs for evaluation; (2) a tumor formation risk assessment step: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the risk of tumor formation; (3) optionally comprising a tumor tissue origin assessment step: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the origin of tumor tissue.

[0007] On the one hand, the present application provides a method for determining differentially methylated region DMR, which includes the step of dividing the differentially methylated region DMR: determining the differentially methylated region DMR based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

[0008] On the one hand, the present application provides a method for evaluating the correlation between a sample to be tested and the risk of tumor formation, which includes a tumor formation risk assessment step: based on the methylation level of the DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the risk of tumor formation, wherein the step of reducing the influence of the subject's age factor on the assessment result is included, and the sample to be tested is derived from the subject.

[0009] On the one hand, the present application provides a method for evaluating the correlation between the sample to be tested and the tumor tissue origin, which includes a tumor tissue origin evaluation step: based on the methylation level of the DMR of the sample to be tested, the correlation between the sample to be tested and the tumor tissue origin is evaluated by a multi-classification method and logistic regression.

[0010] On the one hand, the present application provides a storage medium that records a program that can run the method described in the present application.

[0011] On the one hand, the present application provides a device, which includes the storage medium described in the present application, and the device optionally includes a processor coupled to the storage medium, and the processor is configured to execute based on the program stored in the storage medium to implement the method described in the present application.

[0012] On the one hand, the present application provides a system for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the origin of tumor tissue, characterized in that it comprises: (1) a differential methylation region DMR division module: used to determine multiple target DMRs for evaluation based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites; (2) a tumor formation risk assessment module: used to evaluate the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the target DMR of the sample to be tested; (3) optionally comprising a tumor tissue origin assessment module: used to evaluate the correlation between the sample to be tested and the origin of tumor tissue based on the methylation level of the target DMR of the sample to be tested.

[0013] On the one hand, the present application provides a system for determining differentially methylated region DMRs, characterized in that it comprises a differentially methylated region DMR division module for determining differentially methylated region DMRs based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

[0014] On the one hand, the present application provides a system for evaluating the correlation between a sample to be tested and the risk of tumor formation, characterized in that it comprises a tumor formation risk assessment module: for evaluating the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the DMR of the sample to be tested, wherein it comprises a module for reducing the influence of the subject's age factor on the assessment result, and the sample to be tested is derived from the subject.

[0015] On the one hand, the present application provides a system for evaluating the correlation between a sample to be tested and its tumor tissue origin, characterized in that it comprises a tumor tissue origin evaluation module: used to evaluate the correlation between the sample to be tested and its tumor tissue origin based on the methylation level of the DMR of the sample to be tested, through a multi-classification method and logistic regression.

[0016] This application provides a low-cost, high-precision method that helps to accurately predict and assess the risks of various cancers.

[0017] Those skilled in the art can easily discern other aspects and advantages of the present application from the detailed description below. In the detailed description below, only exemplary embodiments of the present application are shown and described. As will be appreciated by those skilled in the art, the content of this application enables those skilled in the art to modify the disclosed specific embodiments without departing from the spirit and scope of the invention to which this application relates. Accordingly, the descriptions in the drawings and specification of this application are merely exemplary and not restrictive. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The specific features of the inventions of this application are set forth in the appended claims. The features and advantages of the inventions of this application can be better understood by referring to the exemplary embodiments described in detail below and the accompanying drawings. A brief description of the drawings is as follows:

[0019] FIG1 shows an exemplary situation (a theoretical exemplary display, not intended to represent an actual sequencing situation).

[0020] FIG2 shows another exemplary situation (a theoretical exemplary display, not intended to represent an actual sequencing situation).

[0021] 3A-3C show another exemplary situation (a theoretical exemplary display, not intended to represent an actual sequencing situation).

[0022] FIG4 shows that in 5-fold cross validation, a tissue traceability accuracy of 98% (95% CI: 96-99%) could be achieved.

[0023] FIG5 shows the control result of the weight configuration of the Salmon-DOC model of the present application for the confusion-related features.

[0024] [Corrected 23.11.2022 according to Rule 91] Figures 6A-6F show that the Salmon-DOC model of the present application can efficiently detect six cancer types at different stages in the tumor group model.

[0025] [Corrected 23.11.2022 according to Rule 91] Figures 7A-7F show that in the healthy group, the Salmon-DOC model of this application overcomes the previous weakness of increasing false positive methylation with age and maintains balance in all age groups (the horizontal axis is age and the vertical axis is the model cancer probability score).

[0026] 8A-8D show that the traceability accuracy of the Salmon-TOO two-layer model of the present application is better than that of the single-layer model in both cross-validation and independent validation.

[0027] Figure 9 shows the tissue traceability assessment results based on 103 TOO-related DMR regions. DETAILED DESCRIPTION

[0028] The following describes the implementation of the present invention through specific embodiments. People familiar with this technology can easily understand other advantages and effects of the present invention from the contents disclosed in this specification.

[0029] Definition of terms

[0030] In this application, the term "second-generation gene sequencing (NGS)", "high-throughput sequencing" or "next-generation sequencing" generally refers to the second-generation high-throughput sequencing technology and the higher-throughput sequencing methods developed thereafter. Next-generation sequencing platforms include but are not limited to existing sequencing platforms such as Illumina. With the continuous development of sequencing technology, those skilled in the art will understand that other sequencing methods and devices can also be used in this method. For example, second-generation gene sequencing can have the advantages of high sensitivity, high throughput, high sequencing depth, or low cost. According to the development history, influence, sequencing principles and technologies, there are mainly the following types: Massively Parallel Signature Sequencing (MPSS), Polony Sequencing, 454 pyro sequencing, Illumina (Solexa) sequencing, Ion semiconductor sequencing, DNA nano-ball sequencing, Complete Genomics' DNA nanoarray and combined probe anchor ligation sequencing methods, etc. Second-generation gene sequencing enables detailed and comprehensive analysis of a species' transcriptome and genome, hence the name deep sequencing. For example, the methods of this application can also be applied to first-generation gene sequencing, second-generation gene sequencing, third-generation gene sequencing, or single-molecule sequencing (SMS).

[0031] In this application, the term "test sample" generally refers to a sample that needs to be tested. For example, one or more gene regions in the test sample can be tested for the presence of a modification state.

[0032] In this application, the terms "polynucleotide", "nucleotide", "nucleic acid" and "oligonucleotide" are used interchangeably. They refer to polymeric forms of nucleotides (deoxyribonucleotides or ribonucleotides) of any length, or their analogs. Polynucleotides can have any three-dimensional structure and can perform any function, whether known or unknown. The following are non-limiting examples of polynucleotides: coding or non-coding regions of genes or gene fragments, loci (locus) defined by linkage analysis, exons, introns, messenger RNA (mRNA), transfer RNA (tRNA), ribosomal RNA (rRNA), short interfering RNA (siRNA), short-hairpin RNA (shRNA), microRNA (miRNA), ribozymes, cDNA, recombinant polynucleotides, branched polynucleotides, plasmids, vectors, isolated DNA of any sequence, isolated RNA of any sequence, nucleic acid probes, primers and linkers. Polynucleotides can include one or more modified nucleotides, such as methylated nucleotides and nucleotide analogs.

[0033] In this application, the term "methylation" generally refers to the methylation state of a gene fragment, nucleotide, or base thereof in this application. For example, a DNA fragment containing a gene in this application may have methylation on one strand or multiple strands. For example, a DNA fragment containing a gene in this application may have methylation at one site or multiple sites.

[0034] In this application, the term "human reference genome" generally refers to a human genome that can serve as a reference for gene sequencing. Information on the human reference genome can be found at UCSC. The human reference genome can be available in different versions, for example, hg19, GRCH37, or Ensembl 75.

[0035] In this application, the term "machine learning model" generally refers to a system or a collection of program instructions and / or data configured to implement an algorithm, process, or mathematical model. In this application, the algorithm, process, or mathematical model can evaluate and provide a desired output based on a given input. In this application, the parameters of the machine learning model may not be explicitly programmed, and in the traditional sense, the machine learning model may not be explicitly designed to follow specific rules in order to provide a desired output for a given input. For example, the use of the machine learning model may mean that the machine learning model and / or the data structure / set of rules that serves as the machine learning model is trained by a machine learning algorithm.

[0036] In this application, the term "comprising" generally means including the features specifically stated, but not excluding other elements.

[0037] In this application, the term "about" generally refers to a variation within a range of 0.5%-10% above or below the specified value, for example, a variation within a range of 0.5%, 1%, 1.5%, 2%, 2.5%, 3%, 3.5%, 4%, 4.5%, 5%, 5.5%, 6%, 6.5%, 7%, 7.5%, 8%, 8.5%, 9%, 9.5%, or 10% above or below the specified value.

[0038] To detect six cancers with high morbidity and mortality rates, including lung cancer, colorectal cancer, liver cancer, ovarian cancer, pancreatic cancer, and esophageal cancer, this application uses a combination of public databases (TCGA) and internal data mining, using a new algorithm to simultaneously compare genomic methylation variations and spatial locations, screening out a total of 2,536 differentially methylated regions (DMRs) that are highly associated with cancer.

[0039] Summary of the Invention

[0040] On the one hand, the present application provides a method for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the origin of tumor tissue, which may include: (1) a differential methylation region DMR division step: based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites, determining multiple target DMRs for evaluation; (2) a tumor formation risk assessment step: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the risk of tumor formation; (3) optionally including a tumor tissue origin assessment step: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the origin of tumor tissue. For example, the method of the present application for evaluating the correlation between the sample to be tested and the risk of tumor formation and / or the origin of the tumor tissue may include: (1) determining multiple target DMRs for evaluation based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites; (2) evaluating the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the target DMR of the sample to be tested; (3) optionally including evaluating the correlation between the sample to be tested and the origin of the tumor tissue based on the methylation level of the target DMR of the sample to be tested.

[0041] On the one hand, the present application provides a method for determining differentially methylated region DMRs, which may include a differentially methylated region DMR division step: determining the differentially methylated region DMR based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

[0042] On the one hand, the present application provides a method for evaluating the correlation between a sample to be tested and the risk of tumor formation, which includes a tumor formation risk assessment step: based on the methylation level of the DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the risk of tumor formation, wherein the step of reducing the influence of the subject's age factor on the assessment result is included, and the sample to be tested is derived from the subject.

[0043] On the one hand, the present application provides a method for evaluating the correlation between the sample to be tested and the tumor tissue origin, which includes a tumor tissue origin evaluation step: based on the methylation level of the DMR of the sample to be tested, the correlation between the sample to be tested and the tumor tissue origin is evaluated by a multi-classification method and logistic regression.

[0044] For example, the method can include sequencing coverage depth based on the methylation site and the methylation level difference between the methylation site and its adjacent methylation site to determine the DMR. For example, the methylation level difference can refer to the methylation level difference. For example, the methylation level difference can refer to the absolute value of the methylation level difference. For example, the application can determine the DMR region with basically the same methylation level by the methylation level difference between the methylation site and its adjacent methylation site. For example, the application can make the division of the DMR region more accurate by sequencing coverage depth of the methylation site. For example, the data information of the site with higher coverage depth is more reliable.

[0045] For example, the method may include determining the absolute value of the difference in methylation levels between a methylation site and an adjacent methylation site, and determining whether the methylation site and the adjacent methylation site are classified into the same DMR based on the absolute value of the difference. For example, the method may include determining a weight for the absolute value of the difference, wherein the weight for the absolute value of the difference is determined based on the sequencing coverage depth of the methylation site. For example, the weight may be Among them, d ij is the sequencing coverage depth of the jth site in the i-th sample.

[0046] For example, the method may include determining the absolute value of the difference in methylation levels between a methylation site and its adjacent methylation site, and determining the weight of the absolute value of the difference, thereby determining the degree of methylation level difference; the weight of the absolute value of the difference is determined according to the sequencing coverage depth of the methylation site.

[0047] For example, the methylation level difference β ij Determined according to the following formula:

[0048]

[0049] Among them, Mij is the methylation level of the jth site in the i-th sample, e represents the natural constant, P ij Determined according to the following formula:

[0050]

[0051] Among them, d ij is the sequencing coverage depth of the jth site in the i-th sample.

[0052] For example, the methylation level difference β of the methylation site ij When the value is less than or equal to about 0.25, it is determined that the methylation site and its adjacent methylation sites are classified into the same DMR.

[0053] For example, the method may further comprise determining the degree of methylation level fluctuation of the DMR based on the difference between the methylation level difference of the methylation site in the DMR and the methylation site in the middle position of the DMR. For example, the middle position refers to the middle position in the physical position. For example, when M is an odd number and the DMR has M methylation sites, the middle position may refer to the methylation site at about (M+1) / 2 from upstream to downstream. For example, when M is an even number and the DMR has M methylation sites, the middle position may refer to the methylation site at about M / 2 or M / 2+1 from upstream to downstream.

[0054] For example, by determining the degree of fluctuation between the methylation difference between each methylation site in the candidate DMR and the methylation difference between the methylation sites in the middle, a more preferred DMR among the candidate DMRs can be screened.

[0055] For example, the degree of fluctuation of the methylation level of the DMR is β ij Determined according to the following formula:

[0056]

[0057] The β ij is the methylation level difference of the jth site in the i-th sample, μ j is the difference in methylation levels of the methylation sites in the middle of the DMR region. ij A DMR of less than about 1 is used to assess the relevance of the test sample to the risk of tumor formation and / or tumor tissue origin.

[0058] For example, the method may include evaluating the risk of tumor formation in the sample to be tested based on the methylation level of the DMR of the sample to be tested through a binary classification model, and the evaluation method reduces the influence of the subject's age factor on the evaluation results of the correlation between the sample to be tested and the risk of tumor formation and / or the source of the tumor tissue, and the sample to be tested is derived from the subject.

[0059] For example, the binary classification model may include a support vector machine (SVM) model. For example, the method may include introducing a penalty term based on the age factor into the SVM model. For example, the method may include introducing a penalty term based on the age factor into the SVM model using the Hilbert-Schmidt independence criterion. For example, any method for introducing a penalty term that can be used in machine learning in this application can be used in this application to reduce the impact of the age factor.

[0060] For example, the method may include performing machine learning training on training samples known to have tumor formation according to the following formula:

[0061] f(x;w,b)=sgn(wTx+b)

[0062] If a<0, then sgn(a)=-1; if a≥0, then sgn(a)=1

[0063] The following equations are used to determine the training parameters:

[0064]

[0065] Limited by y i (wTx+b)≥1-ξ i

[0066] ξ i ≥0

[0067] C, w, λ, b represent training parameters, sgn() represents the symbol function, ξ i Represents sample x i The degree of violation of the equation, x represents the methylation level of the sample, y represents +1 when the sample is cancerous tissue related to tumor formation, and y represents -1 when the sample is non-cancerous tissue not related to tumor formation, L H (P h(x)h(z) ) is determined by the following formula:

[0068]

[0069]

[0070] h(y) and h(z) are the kernel functions of Y and Z respectively, F and G represent the reproducing kernel Hilbert space of X and Z respectively, P h(x)h(z) represents the probability distribution of h(y) and h(z).

[0071] For example, the method may include determining a classification probability based on the methylation level of the DMR of the test sample using a multi-classification method, and fitting the classification probability using logistic regression to evaluate the correlation between the test sample and the tumor tissue source. For example, the method may determine the classification probability using pairwise voting. For example, the method may determine the classification probability using various multi-classification methods known in the art. For example, the method may fit the classification probability using multivariate linear regression (MLR).

[0072] For example, the method may include performing a regression analysis on training samples of known tissue origin according to the following formula:

[0073] Classification probability determined by binary classification Determined according to the following formula:

[0074]

[0075] μ ij ≡P(y=i|y=i or j,x)

[0076] Where I(x) is the objective equation: if x is true then I {x} =1, if x is false then I {x} =-1, r ij is the pairwise classification probability μ ij , k is the total number of tissue categories; i and j represent the i-th and j-th categories respectively, and x represents the methylation level of the DMR of the sample;

[0077] Weight β of multiple linear regression MLR fitting j Determined according to the following formula:

[0078]

[0079] Among them, X′ i Indicates the classification probability obtained by binary voting, Y ij Indicates the tissue category of the sample.

[0080] For example, the method may correct the tissue origin of the training sample based on the probability of the sample having tumor formation. For example, the method may include performing the correction before obtaining the classification probability through pairwise binary voting. For example, the method may include performing the correction after obtaining the classification probability through pairwise binary voting and before performing the multivariate linear regression analysis. For example, the method may include performing the correction based on a quasi-maximum likelihood estimation method.

[0081] For example, the method may comprise performing a correction according to the following formula:

[0082]

[0083] y i Indicates the organizational source category of the sample, w i represents the weight of the correction, π i represents the probability that the sample has a tumor. For example, by maximizing the expected value of this formula, weights are determined so that the tissue origin category can be corrected based on whether the sample has a tumor. For example, when evaluating a sample with a tumor, information about its tissue origin can be more reliable.

[0084] On the one hand, the present application provides a system for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the origin of tumor tissue, which may include: (1) a differential methylation region DMR division module: based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites, determining multiple target DMRs for evaluation; (2) a tumor formation risk assessment module: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the risk of tumor formation; (3) optionally including a tumor tissue origin assessment module: based on the methylation level of the target DMR of the sample to be tested, evaluating the correlation between the sample to be tested and the origin of tumor tissue.

[0085] On the one hand, the present application provides a system for determining differentially methylated region DMRs, which may include a differentially methylated region DMR division module: determining differentially methylated region DMRs based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

[0086] On the one hand, the present application provides a system for evaluating the correlation between a sample to be tested and the risk of tumor formation, which includes a tumor formation risk assessment module: based on the methylation level of the DMR of the sample to be tested, the correlation between the sample to be tested and the risk of tumor formation is evaluated, wherein, it includes a module for reducing the influence of the subject's age factor on the assessment result, and the sample to be tested is derived from the subject.

[0087] On the one hand, the present application provides a system for evaluating the correlation between the sample to be tested and the tumor tissue origin, which includes a tumor tissue origin evaluation module: based on the methylation level of the DMR of the sample to be tested, the correlation between the sample to be tested and the tumor tissue origin is evaluated through a multi-classification method module and logistic regression.

[0088] For example, the system can include the sequencing coverage depth based on the methylation site and the methylation level difference between the methylation site and its adjacent methylation site to determine the DMR. For example, the methylation level difference can refer to the methylation level difference. For example, the methylation level difference can refer to the absolute value of the methylation level difference. For example, the application can determine the DMR region with basically the same methylation level by the methylation level difference between the methylation site and its adjacent methylation site. For example, the application can make the division of the DMR region more accurate by the sequencing coverage depth of the methylation site. For example, the data information of the site with higher coverage depth is more reliable.

[0089] For example, the system may include determining the absolute value of the difference in methylation levels between a methylation site and its adjacent methylation site, and determining whether the methylation site and its adjacent methylation site are classified into the same DMR based on the absolute value of the difference. For example, the system may include determining a weight for the absolute value of the difference, and the weight for the absolute value of the difference is determined based on the sequencing coverage depth of the methylation site. For example, the weight may be Among them, d ij is the sequencing coverage depth of the jth site in the i-th sample.

[0090] For example, the system may include determining the absolute value of the difference in methylation levels between a methylation site and its adjacent methylation site, and determining the weight of the absolute value of the difference, thereby determining the degree of methylation level difference; the weight of the absolute value of the difference is determined based on the sequencing coverage depth of the methylation site.

[0091] For example, the methylation level difference β ij Determined according to the following formula:

[0092]

[0093] Among them, M ij is the methylation level of the jth site in the i-th sample, e represents the natural constant, P ij Determined according to the following formula:

[0094]

[0095] Among them, d ij is the sequencing coverage depth of the jth site in the i-th sample.

[0096] For example, the methylation level difference β of the methylation site ij When the value is less than or equal to about 0.25, it is determined that the methylation site and its adjacent methylation sites are classified into the same DMR.

[0097] For example, the system may further include determining the degree of methylation level fluctuation of the DMR based on the difference between the methylation level difference of the methylation site in the DMR and the methylation site in the middle position of the DMR. For example, the middle position refers to the middle position in the physical position. For example, when M is an odd number and the DMR has M methylation sites, the middle position may refer to the methylation site at about (M+1) / 2 from upstream to downstream. For example, when M is an even number and the DMR has M methylation sites, the middle position may refer to the methylation site at about M / 2 or M / 2+1 from upstream to downstream.

[0098] For example, by determining the degree of fluctuation between the methylation difference between each methylation site in the candidate DMR and the methylation difference between the methylation sites in the middle, a more preferred DMR among the candidate DMRs can be screened.

[0099] For example, the methylation level fluctuation degree B of the DMR ij Determined according to the following formula:

[0100]

[0101] The β ij is the methylation level difference of the jth site in the i-th sample, μ j is the difference in methylation levels of the methylation sites in the middle of the DMR region. ij A DMR of less than about 1 is used to assess the relevance of the test sample to the risk of tumor formation and / or tumor tissue origin.

[0102] For example, the system may include evaluating the risk of tumor formation in the sample to be tested based on the methylation level of the DMR of the sample to be tested through a binary classification model, and the evaluation system reduces the influence of the subject's age factor on the evaluation results of the correlation between the sample to be tested and the risk of tumor formation and / or the source of the tumor tissue, and the sample to be tested is derived from the subject.

[0103] For example, the binary classification model may include a support vector machine (SVM) model. For example, the system may include introducing a penalty term based on the age factor into the SVM model. For example, the system may include introducing a penalty term based on the age factor into the SVM model using the Hilbert-Schmidt independence criterion. For example, any method for introducing a penalty term that can be used in machine learning in this application can be used in this application to reduce the impact of the age factor.

[0104] For example, the system may include performing machine learning training on training samples known to have tumor formation according to the following formula:

[0105] f(x;w,b)=sgn(wTx+b)

[0106] If a<0, then sgn(a)=-1; if a≥0, then sgn(a)=1

[0107] The following equations are used to determine the training parameters:

[0108]

[0109] Limited by y i (wTx+b)≥1-ξ i

[0110] ξ i ≥0

[0111] C, w, λ, b represent training parameters, sgn() represents the symbol function, ξ i Represents sample x i The degree of violation of the equation, x represents the methylation level of the sample, y represents +1 when the sample is cancerous tissue related to tumor formation, and y represents -1 when the sample is non-cancerous tissue not related to tumor formation, L H (P h(x)h(z) ) is determined by the following formula:

[0112]

[0113]

[0114] h(y) and h(z) are the kernel functions of Y and Z respectively, F and G represent the reproducing kernel Hilbert space of X and Z respectively, P h(x)h(z) represents the probability distribution of h(y) and h(z).

[0115] For example, the system may include determining classification probabilities based on the methylation levels of DMRs in the sample to be tested using a multi-classification method module, and fitting the classification probabilities using logistic regression to assess the correlation between the sample to be tested and the tumor tissue source. For example, the system may determine the classification probabilities using pairwise voting. For example, the system may determine the classification probabilities using various multi-classification method modules known in the art. For example, the system may fit the classification probabilities using multivariate linear regression (MLR).

[0116] For example, the system may include performing a regression analysis on training samples of known tissue origin according to the following formula:

[0117] Classification probability determined by binary classification Determined according to the following formula:

[0118]

[0119] μij ≡P(y=i|y=i or j,x)

[0120] Where I(x) is the objective equation: if x is true then I {x} =1, if x is false then I {x} =-1, r ij is the pairwise classification probability μ ij , k is the total number of tissue categories; i and j represent the i-th and j-th categories respectively, and x represents the methylation level of the DMR of the sample;

[0121] Weight β of multiple linear regression MLR fitting j Determined according to the following formula:

[0122]

[0123] Among them, X′ i Indicates the classification probability obtained by binary voting, Y ij Indicates the tissue category of the sample.

[0124] For example, the system can correct the tissue origin of the training sample based on the probability of the sample forming a tumor. For example, the system can include performing the correction before obtaining the classification probability through pairwise binary voting. For example, the system can include performing the correction after obtaining the classification probability through pairwise binary voting and before performing the multivariate linear regression analysis. For example, the system can include performing the correction based on a quasi-maximum likelihood estimation method module.

[0125] For example, the system may include a correction according to the following formula:

[0126]

[0127] y i Indicates the organizational source category of the sample, w i represents the weight of the correction, π i represents the probability that the sample has a tumor. For example, by maximizing the expected value of this formula, weights are determined so that the tissue origin category can be corrected based on whether the sample has a tumor. For example, when evaluating a sample with a tumor, information about its tissue origin can be more reliable.

[0128] In one aspect, the present application provides a storage medium that records a program that can run the method described in the present application. For example, the non-volatile computer-readable storage medium may include a floppy disk, a flexible disk, a hard disk, a solid-state storage (SSS), or a memory card.

[0129] (e.g., solid-state drive (SSD)), solid-state card (SSC), solid-state module (SSM)), enterprise flash drive, magnetic tape, or any other non-transitory magnetic medium. Non-volatile computer-readable storage media may also include punched cards, paper tape, cursor sheet (or any other physical medium with a hole pattern or other optically identifiable markings), compact disc read-only memory (CD-ROM), compact disc rewritable (CD-RW), digital versatile disc (DVD), Blu-ray disc (BD), and / or any other non-transitory optical medium.

[0130] On the one hand, the present application provides a device, which includes the storage medium described in the present application, and the device optionally includes a processor coupled to the storage medium, and the processor is configured to execute based on the program stored in the storage medium to implement the method described in the present application.

[0131] Example

[0132] Example 1

[0133] The sample is subjected to exemplary bisulfite-treated second-generation sequencing, and the sequencing data obtained include the methylation level and sequencing coverage depth for the methylation site CpG. Optionally, noise removal is performed for the genomic methylation signal CpG and the noise region CHH / CHG site. Then, the p-value obtained by weighted logistic regression is calculated for the "tumor" (C) and "normal" (N) groups. The explanatory variable of the logistic regression adopts a continuous variable, that is, the methylation level of each CpG point, and the response variable adopts a binary output, i.e. (0, 1), corresponding to C and N. The weighted logistic regression distinguishes C and N for each CpG site and tests the null hypothesis that there is no statistical significance between C and N at the CpG site. The weight is determined according to the coverage depth of each CpG site.

[0134] DMR Division

[0135] Based on the methylation level and sequencing coverage depth of the CpG at the methylation site, determine how to divide the various regions of the DMR. Specifically, the methylation level and sequencing coverage depth of the CpG at the methylation site are calculated according to the following formula:

[0136]

[0137]

[0138]

[0139] Here d ij is the effective coverage depth of the jth site of the i-th sample in group C, M ij is the methylation level of the jth site in the i-th sample of group C, and evaluates the similarity of methylation levels at consecutive sites in genomic space. The deeper the coverage depth and the larger the value of parameter P, the higher the similarity of methylation levels between adjacent CpG sites in the same group.

[0140] FIG1 shows an exemplary situation (a theoretical exemplary display, not intended to represent an actual sequencing situation).

[0141] For the first CpG site in the region, samples A and B were covered by 500 valid sequences respectively, and sample C was covered by 200 valid sequences. For sample A, the methylation level of this CpG site was 0.2. The methylation level of the second CpG site in sample A was 0. The coverage depth parameter value P of the first CpG site in this group was calculated for the three samples and was 0.617. At this time, β ij =|0.2-0|*e (1-0.617) =0.29. At the same time, considering that the difference in methylation levels between the two adjacent CpG sites is less than 0.25, which is one of the necessary conditions for classifying the two adjacent sites into the same DMR, the first and second CpG sites in this example will not be classified into the same DMR.

[0142] FIG2 shows another exemplary situation (a theoretical exemplary display, not intended to represent an actual sequencing situation).

[0143] If the above samples are replaced with A, B, and D (where sample D has 400 valid sequences covering the first CpG site), similarly, for sample A, the methylation level of this CpG site is 0.2. The methylation level of the second CpG site of sample A is 0. However, due to the increased sequencing coverage depth of sample D in this example, the coverage depth parameter value P of the first CpG site of this group calculated for the three samples is 0.962. At this time, β ij =|0.2-0|*e (1-0.962) =0.21, which is less than the threshold of 0.25 for classification into the same DMR. At this time, according to sample A, the first and second CpG sites in this example meet the prerequisite for being classified into the same DMR.

[0144] Therefore, the coverage depth of CpG sites introduced by the method of the present application can significantly improve the accuracy of DMR region division.

[0145] Further optionally, for B in a region ij , calculated as follows

[0146]

[0147]

[0148] Figures 3A-3C show another exemplary situation (a theoretical exemplary display, not used to represent the actual sequencing situation). When the DMR region contains 10 CpG sites, the B ij The scores of each DMR were calculated by merging and averaging.

[0149] The steps for calculating the B value in the DMR region shown in group A are shown in the following table:

[0150]

[0151] B is worth 0.1, that is

[0152] Similarly, the B-value score within the DMR shown in group B is 0.7, that is, The B value score in the DMR shown in group C is 1.233, that is,

[0153] The DMR regions screened by this method not only contain cancer mutation information of various cancer types, but also contain tissue-specific characteristics, and have better segmentation effects at the regional boundaries.

[0154] Figure 4 shows that for six types of cancer: lung cancer (LC), colorectal cancer (CRC), liver cancer (LIHC), ovarian cancer (OVCA), pancreatic cancer (PAAD), and esophageal cancer (ESCA), a tissue traceability accuracy of 98% (95% CI: 96-99%) can be achieved in 5-fold cross validation.

[0155] Example 2

[0156] Cancer Assessment (DOC) model establishment

[0157] The ctDNA content in the blood varies greatly at different stages of development of different cancers and is easily affected by experimental batch effects. In addition, methylation variation is related to age, disease, race, etc. If these are not handled, they may affect the accuracy of the classification model as confounding variables. This application adopts a model construction method called Salmon, which first quantifies the bias caused by the confounding variable (the quantification method can be used but is not limited to the Hilbert-Schmidt independence criterion), and then embeds the regularization term of the model to correct it, thereby increasing the accuracy and generalization ability of the model.

[0158] Algorithm establishment

[0159] Assume m samples, set the feature vector X(x1,…,x m ), classification labels Y(y1,…,y m ), the confusion variables Z(z1,…,z m ), where x i is an n-dimensional vector representing the methylation characteristics of sample i, y i is x i The classification label, y i ∈{-1,+1},z i is some kind of confounding variable for sample i.

[0160]

[0161] Here L H Refers to the Hilbert-Schmidt independence criterion, which is used to measure the degree of independence of variables X and Z. h(y) and h(z) are the kernel functions of Y and Z. P h(x)h(z) represents the probability distribution of h(y) and h(z), F and G represent the reproducing kernel Hilbert space of X and Z respectively, which can be understood as the domain mapped after nonlinear processing of X and Z, C h(x)h(z) Refers to the correlation coefficient of the two kernel functions, and HS is the Hilbert Space.

[0162] ‖C h(y)h(z) ‖ 2 =(E h(x)h(z) -E h(x) E h(z) ) 2 =(E h(x)h(z) )2 +(E h(x) E h(z) ) 2 -2E h(x)h(z) E h(x) E h(z)

[0163] Support vector machine (SVM) is used as the main classifier

[0164] f(x;w,b)=sgn(wTx+b)

[0165] sgn(a)=1(-1)if a≥0(<0)

[0166] The classification interface is determined by solving the following objective equation:

[0167]

[0168] sty i (wTx+b)≥1

[0169] For inseparable data, the soft-margin support vector machine (soft-margin SVM) introduces a penalty term for training errors

[0170]

[0171] sty i (wTx+b)≥1-ξ i

[0172] ξ i ≥0

[0173] Here C controls the balance between minimizing the training error and maximizing the classification margin, while ξ i Refers to sample x i The degree of violation of the equation.

[0174] In order to control the confounding factors, Salmon adds a regularization term to the objective equation solved by SVM. The parameter λ controls the balance between the confounding factor error and the maximum margin width during training. The objective equation is:

[0175]

[0176] sty i (wTx+b)≥1-ξ i

[0177] ξ i ≥0

[0178] Here C and λ control the balance between minimizing the training error, minimizing the correlation between the confounding variables and the explanatory variables, and maximizing the classification interval.

[0179] FIG5 shows the control result of the weight configuration of the Salmon-DOC model of the present application for the confusion-related features.

[0180] Each data point represents a blood sample used in the Salmon-DOC model. The horizontal axis represents the confunding factor for the corresponding sample, and the vertical axes represent the original uncorrected variable coeffect (Panel A) and the corrected variable coeffect (Panel B). Comparison before and after correction demonstrates that confounding-related features are weighted appropriately in the Salmon-DOC model.

[0181] Review of cohort data

[0182] This application uses retrospective clinical samples of 6 cancer types, divided into training sets and validation sets, to evaluate the accuracy of Salmon's binary classifier (cancer vs. non-cancerous).

[0183] FIG6 shows that the Salmon-DOC model of the present application can efficiently detect six cancer types at different stages in the tumor group model.

[0184] Figure 7 shows that in the healthy group, the Salmon-DOC model of this application overcomes the previous weakness of increasing false positive methylation with age and maintains balance in various age groups (the horizontal axis is age, and the vertical axis is the model cancer probability score).

[0185] Example 3

[0186] Establishment of Traceability of Organization (TOO) model

[0187] First-layer TOO model construction

[0188] The TOO model is essentially a multi-classification problem. Calculating the probability of each class can be simplified to voting on paired binary classification results and selecting the one with the most votes. However, for the potential clinical application of tissue provenance models, simply generating a single classification result is insufficient; only by generating the probability of the classification can the model be assembled.

[0189] Therefore, the first step of the Salmon-TOO model in this application is to quantify the binary voting results. This quantification can be proved by probability calculation. If we define a data point x and a label y, we assume that the paired classification probability μij exists, then from the i-th and j-th categories in the training set, we can get a model that can be used to calculate r by inputting any new data point x. ij As μ ij The problem can be simplified to use all r ij To estimate the probability of the i-th category

[0190] p i =P(y=i|x),i=1,…,k

[0191] Define r ij μ ij Assuming μ ij +μ ji =1. For multi-classification problems, a "voting" system is used.

[0192] μ ij ≡P(y=i|y=i or j,x)

[0193]

[0194] Define I as the objective equation: I {x} = 1 if x is true, otherwise false. The probability calculation can be written as

[0195]

[0196] Second layer TOO model construction

[0197] The second layer of the Salmon-TOO model is to perform MLR fitting for different classes

[0198] Assuming a probability calculation for tissue origin is required, the first layer yields quantified binary probabilities, ranging from (∞, -∞). Because the actual distribution of each pair of binary probabilities is inconsistent, the quantified binary probabilities can be further used as explanatory variables in logistic regression, with the response variable taking a multivariate output, corresponding to the known tissue origins during the modeling process.

[0199]

[0200] As shown in the table above, each column represents a feature variable of logistic regression That is, the binary classification evaluation probability of each two organizational categories; each row represents a response variable y1, that is, the organizational category (class).

[0201] Assume that there are J discontinuous reflection variables, which are used to explain the characteristic variables of the binary classification. Then, the evaluation results are converted into Y i1 ,…,Y iJ, β j is the feature weight based on each reflection variable.

[0202]

[0203] In the Salmon-DOC model, we can find that some cancer types are judged as negative, while others are judged as positive. Therefore, in view of this judgment, when conducting traceability modeling, the weight correction of tissue class is performed based on the quasi-maximum likelihood estimation method. Taking binary logistic regression as an example, it can be explained as follows:

[0204]

[0205] Review of cohort data

[0206] The entire data set of the retrospective cohort was randomly split 1:1 into training and validation sets. First, the training set was used to perform cross-validation to generate the traceability assessment results. During this process, the model parameters were continuously optimized and ultimately locked. Finally, the locked model was used to evaluate the traceability results of the entire validation set. The traceability model training set included 300 cases of six cancer types, with a relatively balanced number of cases across each cancer type and stage: 36 cases of lung cancer (4, 12, 5, and 15 cases for stages I to IV, respectively); 62 cases of colorectal cancer (8, 18, 18, and 18 cases for stages I to IV, respectively); 74 cases of liver cancer (25, 14, 22, and 13 cases for stages I to IV, respectively); 48 cases of ovarian cancer (1, 4, 38, and 5 cases for stages I to IV, respectively); 40 cases of pancreatic cancer (3, 6, 13, and 18 cases for stages I to IV, respectively); and 42 cases of esophageal cancer (5, 10, 15, and 12 cases for stages I to IV, respectively). The traceability model validation set consists of 224 samples, including: 31 cases of lung cancer (4 / 5 / 12 / 10 cases in stages I to IV, respectively), 52 cases of colorectal cancer (7 / 15 / 13 / 17 cases in stages I to IV, respectively), 55 cases of liver cancer (17 / 11 / 20 / 7 cases in stages I to IV, respectively), 27 cases of ovarian cancer (3 / 4 / 8 / 12 cases in stages I to IV, respectively), 25 cases of pancreatic cancer (4 / 6 / 6 / 9 cases in stages I to IV, respectively), and 34 cases of esophageal cancer (4 / 7 / 8 / 15 cases in stages I to IV, respectively).

[0207] 8A-8D show that the traceability accuracy of the Salmon-TOO two-layer model of the present application is better than that of the single-layer model in both cross-validation and independent validation.

[0208] Figures 8A and 8B are the traceability evaluation results of the cross-validation of the six cancer types in the six cancer training set. Among them, Figure 8A is the result output after only the first-layer TOO model was constructed, and the traceability accuracy was 0.87 (260 / 300). If the suboptimal traceability result was included, the accuracy was 0.93 (279 / 300); Figure 8B is the output result after the second-layer MLR model was supplemented on the basis of the first-layer TOO model. The traceability accuracy was improved to 0.90 (270 / 300). If the suboptimal traceability result was included, the accuracy could be further improved to 0.95 (284 / 300). Similarly, Figures 8C and 8D are the traceability evaluation results of the six cancer types independently verified in the above-mentioned validation set. Among them, Figure 8C shows the output result after only constructing the first-layer TOO model. The traceability accuracy is 0.77 (173 / 224). If the suboptimal traceability result is included, the accuracy is 0.87 (194 / 224); Figure 8D shows the output result after supplementing the second-layer MLR model on the basis of the first-layer TOO model. The traceability accuracy is improved to 0.84 (187 / 224). If the suboptimal traceability result is included, the accuracy can be further improved to 0.89 (199 / 224).

[0209] In summary, the evaluation accuracy of the Salmon-TOO two-layer traceability model of this application is better than that of the single-layer model in both training set cross-validation and independent validation.

[0210] Example 4

[0211] DOC cancer detection model

[0212] Table 1A shows the 94 DMR regions used in the DOC cancer detection model.

[0213]

[0214]

[0215] Based on 94 DOC-related DMR regions, we evaluated 100 healthy individuals and 318 six-cancer positive samples from the independent validation set 1. The overall sensitivity was 80.5% (256 / 318) and the overall specificity was 95% (95 / 100). While maintaining a specificity of 90%, the sensitivity for specific cancer types and stages is shown in the following table:

[0216]

[0217]

[0218] Repeated testing was then performed, with each test taking 50 random samples from the 94 DOC regions. Maintaining a specificity of 90% (90 / 100), the sensitivity of the six cancer-positive samples in five repeated tests is shown in the following table:

[0219]

[0220] Example 5

[0221] TOO Organization Traceability Model

[0222] Table 1B shows the 103 DMR regions used in the TOO tissue traceability model.

[0223]

[0224]

[0225] Based on 103 TOO-related DMR regions, 473 six-cancer positive samples in the independent validation set 2 were traced and evaluated. The first traceability accuracy was 63.0% (298 / 473). If the suboptimal traceability results were included, the accuracy could be improved to 71.5% (338 / 473).

[0226] Figure 9 shows the tissue traceability assessment results based on 103 TOO-related DMR regions.

[0227] Four rounds of repeated testing were then conducted, each time using 50 random samples from the 103 TOO regions. The traceability accuracy results from the four rounds of evaluation are shown in the following table:

[0228]

[0229] Example 6

[0230] Results of simultaneous DOC and TOO assessments of 222 DMRs:

[0231] Table 1C shows the 222 DMR regions used in the DOC and TOO assessment models.

[0232]

[0233]

[0234]

[0235] In an independent validation set, we calculated the sensitivity and tissue-based accuracy of 473 negative and 473 positive samples of six cancers, using 222 markers, at a uniform specificity of 95.1% (450 / 473). The evaluated tumor detection and tissue-based results are shown in the following table:

[0236]

[0237]

[0238] The foregoing detailed description is provided by way of explanation and example and is not intended to limit the scope of the appended claims. Various changes to the embodiments listed in the present application are obvious to those skilled in the art and are intended to be within the scope of the appended claims and their equivalents.

Claims

1. A method for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the source of tumor tissue, characterized in that: The method comprises: (1) a differentially methylated region DMR division step: determining a plurality of target DMRs for evaluation based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites; (2) Tumor formation risk assessment step: based on the methylation level of the target DMR in the sample to be tested, assessing the correlation between the sample to be tested and the risk of tumor formation; (3) Optionally comprising a tumor tissue origin assessment step: based on the methylation level of the target DMR of the sample to be tested, assessing the correlation between the sample to be tested and the tumor tissue origin.

2. A method for determining a differentially methylated region DMR, characterized in that: The method comprises a differentially methylated region DMR division step: determining the differentially methylated region DMR based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

3. A method for evaluating the correlation between a sample to be tested and the risk of tumor formation, characterized in that: The method comprises a step of assessing the risk of tumor formation: assessing the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the DMR of the sample to be tested, wherein the method comprises a step of reducing the influence of the age factor of the subject on the assessment result, and the sample to be tested is derived from the subject.

4. A method for evaluating the correlation between a sample to be tested and a tumor tissue source, characterized in that: The method comprises a tumor tissue origin assessment step: based on the methylation level of the DMR of the sample to be tested, the correlation between the sample to be tested and the tumor tissue origin is assessed by multi-classification method and logistic regression.

5. The method of any one of claims 1-4, comprising determining the absolute value of the difference in methylation levels between a methylation site and its adjacent methylation site, and determining whether the methylation site and its adjacent methylation site are classified into the same DMR based on the absolute value of the difference.

6. The method of claim 5, comprising determining a weight for the absolute value of the difference, wherein the weight for the absolute value of the difference is determined according to the sequencing coverage depth of the methylation site.

7. The method according to any one of claims 1 to 6, wherein the methylation level difference β ij Determined according to the following formula: in, M ij is the methylation level of the jth site in the ith sample, e represents the natural constant, P ij Determined according to the following formula: Among them, d ij is the sequencing coverage depth of the jth site in the ith sample.

8. The method of claim 7, wherein the methylation level difference of the methylation site is β ij When the value is less than or equal to about 0.25, it is determined that the methylation site and its adjacent methylation site are classified into the same DMR.

9. The method of any one of claims 1-8, further comprising determining the degree of fluctuation of the methylation level of the DMR based on the difference between the methylation level difference of the methylation site in the DMR and the methylation site in the middle position of the DMR.

10. The method of claim 9, wherein the methylation level fluctuation degree of the DMR is ij Determined according to the following formula: The β ij is the methylation level difference of the jth site in the i-th sample, μ j It is the difference in methylation levels of the methylation sites in the middle of the DMR region.

11. The method of claim 10, wherein B ij A DMR of less than about 1 is used to assess the correlation between the sample to be tested and the risk of tumor formation and / or the tumor tissue origin.

12. The method according to any one of claims 1-11, comprising evaluating the correlation between the test sample and the risk of tumor formation through a binary classification model based on the methylation level of the DMR of the test sample, wherein the evaluation method reduces the influence of the subject's age factor on the evaluation result of the correlation between the test sample and the risk of tumor formation and / or the source of tumor tissue, and the test sample is derived from the subject.

13. The method of claim 12, wherein the binary classification model comprises a support vector machine (SVM) model.

14. The method of claim 13, comprising introducing a penalty term based on the age factor into the SVM model.

15. The method of claim 14, comprising introducing a penalty term based on the age factor in the SVM model in the manner of a Hilbert-Schmidt independence criterion.

16. The method of any one of claims 1-15, comprising performing machine learning training on training samples known to have tumor formation or known to have no tumor formation according to the following formula: f(x;w,b)=sgn(wTx+b) If a<0, then sgn(a)=-1 If a≥0, then sgn(a)=1 The following equation is used to determine the training parameters: Limited by y i (wTx+b)≥1-ξ i x i ≥0 C, w, λ, b represent training parameters, sgn() represents the symbol function, ξ i Represents sample x i The degree of violation of the equation, x represents the methylation level of the sample, y represents +1 when the sample is related to tumor formation, and y represents -1 when the sample is not related to tumor formation, L H (P h(x)h(z) ) is determined by the following formula: h(y) and h(z) are the kernel functions of Y and Z, respectively. F and G represent the reproducing kernel Hilbert spaces of X and Z, respectively. h(x)h(z) represents the probability distribution of h(y) and h(z).

17. The method of any one of claims 1-16, comprising determining the classification probability based on the methylation level of the DMR of the sample to be tested by a multi-classification method, and fitting the classification probability by logistic regression to evaluate the correlation between the sample to be tested and the tumor tissue source.

18. The method of claim 17, wherein the classification probability is determined by pairwise voting.

19. The method according to any one of claims 17-18, wherein the classification probability is fitted by multivariate linear regression (MLR).

20. The method of any one of claims 1-19, comprising performing a regression analysis on training samples of known tissue origin according to the following formula: Classification probability determined by two categories Determined according to the following formula: μ ij ≡P(y = i | y = i or j, x) where I(x) is the objective equation: if x is true then I {x} =1, if x is false then I {x} =-1, r ij is the pairwise classification probability μ ij , k is the total number of tissue categories; i and j represent the i-th and j-th categories respectively, and x represents the methylation level of the DMR of the sample; Weight β of multivariate linear regression MLR fitting j Determined according to the following formula: in, X′ i Indicates the classification probability obtained by binary voting, Y ij Indicates the tissue category of the sample.

21. The method of any one of claims 1-20, wherein the tissue origin of the training samples is corrected based on the probability that the samples have tumor formation.

22. The method of claim 21, comprising performing the correction after obtaining the classification probability through the pairwise binary voting and before the multivariate linear regression analysis.

23. The method of any one of claims 21-22, comprising performing the correction based on a quasi-maximum likelihood estimation method.

24. The method of any one of claims 21-23, comprising performing a correction according to the following formula: y i Indicates the organizational source category of the sample, w i represents the weight of the correction, π i Indicates the probability of tumor formation in the sample.

25. A storage medium recording a program capable of executing the method according to any one of claims 1 to 24.

26. A device, comprising the storage medium of claim 25, and optionally comprising a processor coupled to the storage medium, the processor being configured to execute based on a program stored in the storage medium to implement the method of any one of claims 1-24.

27. A system for evaluating the correlation between a sample to be tested and the risk of tumor formation and / or the source of tumor tissue, characterized in that: It includes: (1) differential methylation region DMR division module: used to determine multiple target DMRs for evaluation based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites; (2) a tumor formation risk assessment module: used to assess the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the target DMR in the sample to be tested; (3) Optionally comprising a tumor tissue origin assessment module: used to assess the correlation between the sample to be tested and the tumor tissue origin based on the methylation level of the target DMR of the sample to be tested.

28. A system for determining differentially methylated regions (DMRs), characterized in that: Contains a differentially methylated region DMR division module: used to determine the differentially methylated region DMR based on the sequencing coverage depth of the methylation site and / or the difference in methylation levels of adjacent methylation sites.

29. A system for evaluating the correlation between a sample to be tested and the risk of tumor formation, characterized in that: It comprises a tumor formation risk assessment module: used to assess the correlation between the sample to be tested and the risk of tumor formation based on the methylation level of the DMR in the sample to be tested, wherein it comprises a module for reducing the influence of the age factor of the subject on the assessment result, and the sample to be tested is derived from the subject.

30. A system for evaluating the correlation between a sample to be tested and a tumor tissue source, characterized in that: It comprises a tumor tissue origin assessment module: it is used to assess the correlation between the sample to be tested and the tumor tissue origin based on the methylation level of the DMR of the sample to be tested, through a multi-classification method and logistic regression.

Images