Circulating tumor DNA prediction method based on DNA methylation characteristics
By selecting stable but significantly different CpG sites in the blood, combining the maximum expectation algorithm and likelihood ratio test, predicting ctDNA fragments in cfDNA samples, the problem of difficulty in identifying and capturing ctDNA signals in the prior art is solved, and the accuracy and efficiency of cancer screening are improved.
Patent Information
- Application Number
- CN202311535288.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-17
- Publication Date
- 2025-03-04
Smart Images
Figure CN119560024A_ABST
Abstract
Description
Technical Field
[0001] The present application belongs to the technical field of bioinformatics, and specifically relates to a circulating tumor DNA prediction method based on DNA methylation characteristics. Technical Background
[0002] DNA methylation is an epigenetic modification that not only plays an important role in transcriptional regulation, but is also associated with disease phenotypes. More and more studies have shown that disordered DNA methylation patterns play an important role in the occurrence and development of cancer. During the occurrence and development of cancer, abnormal DNA methylation patterns occur not only in CpG islands near gene promoters, but also in conserved sequences thousands of bases away from the promoter, and the frequency of occurrence is higher. More importantly, abnormal DNA methylation patterns occur in the early stages of cancer, which helps detect early cancer. Therefore, abnormal DNA methylation patterns are one of the hallmarks of cancer and exist in all stages of cancer.
[0003] Cell-free DNA (cfDNA) in the blood comes not only from normal cells, but also from disease-specific apoptotic cells, such as circulating tumor DNA (ctDNA) in cfDNA. At present, cfDNA has become the clinical standard for prenatal non-invasive testing and is used in the field of cancer screening. However, for most cancer patients (especially those with early cancer), the content of ctDNA in cfDNA is very low. Therefore, a challenging problem for cancer screening based on cfDNA is how to effectively capture trace ctDNA signals. At present, high-depth sequencing methods are mainly used to enhance ctDNA signals, but the main problem of this processing method is high cost and difficult to implement in large-scale early screening. In addition, this processing method will also enhance the inherent noise in sequencing, and there is a risk of increasing the false positive rate. A potentially feasible strategy to enhance ctDNA signals is to select reads that may come from tumor cells from cfDNA, and then extract tumor features from the selected reads for early tumor screening. Studies have preliminarily shown that this strategy has certain feasibility. One of the key issues that the above strategy needs to solve is how to identify the cell source of cfDNA in the blood. Since DNA methylation patterns are cell-specific, they can be used to deconstruct the composition of cfDNA, that is, what cell type each read in cfDNA comes from. However, there are thousands of cell types in the human body, and current deconstruction methods based on DNA methylation patterns have failed to effectively identify tumor cell types from thousands of cell types.
[0004] In view of this, this application is filed. Summary of the invention
[0005] In order to solve the above technical problems, the present application selects CpG sites whose DNA methylation levels are stable between samples of the same type but significantly different between samples of different types based on healthy cfDNA samples, chronic inflammation cfDNA samples and tumor tissue samples; for the selected CpG sites, a statistical model is constructed based on the reference data set (healthy cfDNA samples, chronic inflammation cfDNA samples and tumor tissue samples) and the detected cfDNA samples, and the Expectation-Maximization algorithm (EM) algorithm is used to estimate the methylation levels of the selected CpG sites in samples of different types and the proportion of ctDNA in the detected cfDNA samples; based on the estimation results of the EM algorithm, the Likelihood Ratio Test (LRT) is used to determine the sample type of the detected cfDNA sample, that is, whether it is a healthy sample, a chronic inflammation sample or a tumor sample; according to the likelihood ratio test results, the ctDNA fragments in the detected cfDNA samples are predicted based on the estimated methylation levels of the CpG sites and the proportion of ctDNA.
[0006] Therefore, the core purpose of this application is to provide a circulating tumor DNA prediction method based on DNA methylation characteristics.
[0007] To achieve the above objectives, this application proposes the following technical solutions:
[0008] 1) Select CpG sites whose DNA methylation levels are stable among samples of the same type but significantly different among samples of different types;
[0009] 2) Construct a statistical model and use the expectation-maximization algprithm (EM) algorithm to estimate the methylation levels of the selected CpG sites in different types of samples and the proportion of ctDNA in the detected cfDNA samples;
[0010] 3) Use the likelihood ratio test (LRT) to determine the sample type of the cfDNA sample tested;
[0011] 4) Predict and detect ctDNA fragments in cfDNA samples.
[0012] Furthermore, the selection in step 1) is: based on healthy cfDNA samples, chronic inflammation cfDNA samples and tumor tissue samples, a binomial distribution is constructed to select CpG sites whose DNA methylation levels are stable between samples of the same type but significantly different between samples of different types.
[0013] Furthermore, the selection is specifically as follows: a. Based on healthy cfDNA samples, a binomial distribution is constructed for the DNA methylation level of each CpG site and whether it is stable in the healthy cfDNA samples; b. Based on chronic inflammation cfDNA samples, a binomial distribution is constructed for the DNA methylation level of each CpG site that is stable in the healthy cfDNA samples and whether it is stable in the chronic inflammation cfDNA samples and whether there is a significant difference between the healthy and chronic inflammation samples; c. Based on tumor tissue samples, a binomial distribution is constructed for the DNA methylation level of each CpG site that is stable in the chronic inflammation cfDNA samples and significantly different between the healthy and chronic inflammation samples and whether it is stable in the tumor tissue samples and whether there is a significant difference between the three samples of tumor, healthy samples and chronic inflammation.
[0014] Furthermore, the construction of step 2) is as follows: the DNA fragments in cfDNA are divided into two types: tumor DNA and non-tumor DNA, and a statistical model M1 is constructed based on the CpG sites selected in step 1).
[0015] Furthermore, the statistical model M1 is:
[0016] P(X,Y|β,θ)=P(X|β,θ)P(Y|β)
[0017]
[0018]
[0019]
[0020]
[0021]
[0022] y i1 , y' i1 : The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m
[0023] y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m
[0024] x i , x' i : The number of methylation reads and total reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m
[0025] β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m
[0026] θ: Detect the proportion of ctDNA in cfDNA samples
[0027] Furthermore, the expectation-maximization algorithm (EM) is used to solve the statistical model M1 constructed in step 2, as follows:
[0028] Initial step (I-step):
[0029] N is the number of sequencing fragments in the cfDNA sample
[0030]
[0031] Expected step (E-step):
[0032]
[0033]
[0034] Maximization step (M-step):
[0035]
[0036]
[0037] Furthermore, the alternative hypothesis model of the likelihood ratio test in step 3) (detecting that cfDNA samples contain ctDNA fragments) is the M1 model of step 2), and the null hypothesis model (detecting that cfDNA samples do not contain ctDNA fragments) is:
[0038] P(X, Y|β)=P(X|β)P(Y|β)
[0039]
[0040]
[0041]
[0042]
[0043] y i1 , y' i1: The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m
[0044] y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m
[0045] x i , x' i : The number of methylation reads and total reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m
[0046] β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m
[0047] Furthermore, the likelihood ratio test statistic based on the aforementioned M0 model and M1 model follows a chi-square distribution with 1 degree of freedom and a significance level of 0.05.
[0048] Furthermore, the prediction of step 4) is: based on the likelihood ratio test result of step 3), if the null hypothesis is not true, then according to the methylation level estimated in step 2) and the proportion of ctDNA in the detected cfDNA sample, calculate the probability that fragment c in the detected cfDNA sample comes from tumor tissue, and then determine whether fragment c comes from tumor tissue by using a random number generated by a pseudo-random number generator.
[0049] Furthermore, the formula for calculating the probability that fragment c in the cfDNA sample comes from tumor tissue is:
[0050]
[0051]
[0052] β i1 , β i2 : Estimated value of methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, …, m
[0053] θ: Estimated value of the proportion of ctDNA in the cfDNA sample
[0054] Beneficial technical effects of this application:
[0055] This application characterizes the DNA methylation patterns of non-tumor fragments and tumor fragments in cfDNA, and performs statistical tests on whether the detected cfDNA samples contain ctDNA, and predicts the ctDNA fragments in the cfDNA samples based on the statistical test results and the estimated ctDNA ratio, thereby enriching the tumor signals in the cfDNA samples, providing the necessary technical support for improving the sensitivity and specificity of early cancer screening. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 , The present application provides a flowchart for predicting circulating tumor DNA based on DNA methylation characteristics. DETAILED DESCRIPTION
[0057] The embodiments of the present application will be described in detail below in conjunction with the examples, but it will be appreciated by those skilled in the art that the following examples are only used to illustrate the present application and should not be considered as limiting the scope of the present application. In the examples, if no specific conditions are specified, the conditions are carried out according to normal conditions or manufacturer recommendations. If the manufacturer is not specified for the reagents or instruments used, they are all conventional products that can be purchased from the market.
[0058] Partial term definition, unless otherwise defined below, the meaning of all technical terms and scientific terms used in the specific embodiments of the present application is intended to be the same as that commonly understood by those skilled in the art. Although it is believed that the following terms are well understood by those skilled in the art, the following definitions are still set forth to better explain the present application.
[0059] The term "approximately" in this application indicates an accuracy range that can be understood by those skilled in the art and still ensures the technical effect of the characteristic in question. The term generally indicates ±10%, preferably ±5%, of the indicated value.
[0060] As used in this application, the terms "comprises", "comprising", "having", "containing" or "involving" are inclusive or open-ended and do not exclude other unrecited elements or method steps. The term "consisting of" is considered a preferred embodiment of the term "comprising". If a group is defined below as comprising at least a certain number of embodiments, this should also be understood to disclose a group that preferably consists of only these embodiments.
[0061] In addition, the terms first, second, third, (a), (b), (c), and the like in the specification and claims are used to distinguish similar elements and are not necessarily required to describe a sequential or chronological order. It should be understood that the terms so used are interchangeable under appropriate circumstances, and the embodiments described in this application can be implemented in other sequences than those described or illustrated in this application.
[0062] The present application is described below in conjunction with specific embodiments.
[0063] Example Prediction Method Establishment
[0064] like Figure 1 As shown, the present application provides a circulating tumor DNA prediction method based on DNA methylation characteristics, comprising the following steps:
[0065] S1. Select CpG sites: Based on healthy cfDNA samples, chronic inflammation cfDNA samples and tumor tissue samples, construct a binomial distribution to select CpG sites whose DNA methylation levels are stable among samples of the same type but significantly different among samples of different types. The specific operations are as follows:
[0066] (1) Based on healthy cfDNA samples, a binomial distribution is constructed for the DNA methylation level of each CpG site and tested whether it is stable in healthy cfDNA samples, including the following four steps:
[0067] a. Based on healthy cfDNA samples, count the number of methylation reads at the i-th CpG site (denoted as x 1i M ) and the number of unmethylated reads (denoted as ), and calculate the methylation rate
[0068] b. For the jth cfDNA sample in the healthy cfDNA sample, count the number of methylation reads of the i-th CpG site (denoted as x 1ij M ) and the number of unmethylated reads (denoted as ), and based on the binomial distribution Calculate the number of methylation of the i-th CpG site in the j-th cfDNA sample (x 1ij M ) occurs, if the obtained probability is less than 0.05, the sample is marked as abnormally methylated at the i-th CpG site;
[0069] c. If the proportion of samples with abnormal methylation at the i-th CpG site in healthy cfDNA samples is greater than or equal to 0.05, the methylation level of the i-th CpG site is marked as unstable;
[0070] d. All CpG sites were marked according to the above method, and the CpG sites marked as unstable were excluded. The remaining CpG sites were regarded as CpG sites with stable methylation levels among healthy cfDNA samples.
[0071] (2) Based on the chronic inflammation cfDNA samples, a binomial distribution was constructed for the DNA methylation level of each CpG site that was stable in the healthy cfDNA samples and tested whether it was stable in the chronic inflammation cfDNA samples and whether there was a significant difference between the healthy and chronic inflammation samples, including the following 6 steps:
[0072] a. For the i-th stable CpG site in the healthy cfDNA sample, count the number of methylation reads in the chronic inflammation cfDNA sample (denoted as x 2i M ) and the number of unmethylated reads (denoted as ), and calculate the methylation rate
[0073] b. For the i-th stable CpG site in the healthy cfDNA sample, count the number of methylation reads in the j-th chronic inflammation cfDNA sample (denoted as x 2ij M ) and the number of unmethylated reads (denoted as ), and based on the binomial distribution Calculate the number of methylation of the CpG site in the jth chronic inflammation cfDNA sample (x 2ij M ) occurs, if the obtained probability is less than 0.05, the sample is marked as abnormally methylated at the CpG site;
[0074] c. For the i-th stable CpG site in the healthy cfDNA sample, if the proportion of samples with abnormal methylation of this CpG site in the chronic inflammation cfDNA samples is greater than or equal to 0.05, it is marked that the methylation level of this CpG site is unstable among the chronic inflammation cfDNA samples;
[0075] d. All stable CpG sites in healthy cfDNA samples were marked according to the above method, and the CpG sites marked as unstable among chronic inflammation cfDNA samples were excluded. The remaining CpG sites were regarded as CpG sites with stable methylation levels among chronic inflammation cfDNA samples;
[0076] e. For the i-th CpG site that is stable in both healthy cfDNA samples and chronic inflammation cfDNA samples, based on the number of methylation reads in the healthy cfDNA sample (denoted as x 1i M ) and the number of unmethylated reads (denoted as ), and the methylation rate p in chronic inflammatory cfDNA samples 2i , according to the binomial distribution Calculate the probability P(x = x 1iM ); if the obtained probability is less than 0.05, it indicates that there is a significant difference in the methylation level of the CpG site between the healthy cfDNA samples and the chronic inflammation cfDNA samples;
[0077] f. All CpG sites that are stable in both healthy cfDNA samples and chronic inflammation cfDNA samples are marked with significant differences, and CpG sites whose methylation levels are significantly different between healthy and chronic inflammation samples but stable in each sample are obtained.
[0078] (3) Based on tumor tissue samples, a binomial distribution was constructed for the DNA methylation level of each CpG site that was stable in chronic inflammation cfDNA samples but significantly different between healthy and chronic inflammation samples, and the results were tested to see whether it was stable in tumor tissue samples and whether there was a significant difference between tumor, healthy and chronic inflammation samples, including the following 7 steps:
[0079] a. For the i-th CpG site that is significantly different between the healthy and chronic inflammation samples but stable in each sample, the number of methylation reads in the tumor tissue sample is counted (denoted as y i M ) and the number of unmethylated reads (denoted as ), and calculate the methylation rate
[0080] b. For the i-th CpG site that is significantly different between the healthy and chronic inflammation samples but stable in each sample, the number of methylation reads in the j-th tumor tissue sample (denoted as y ij M ) and the number of unmethylated reads (denoted as ), and based on the binomial distribution Calculate the number of times the CpG site in the jth tumor tissue sample is methylated (y ij M ) occurs, if the obtained probability is less than 0.05, the sample is marked as abnormally methylated at the CpG site;
[0081] c. For the i-th CpG site that is significantly different between the healthy and chronic inflammation samples but stable in each sample, if the proportion of samples with abnormal methylation of this CpG site in tumor tissue samples is greater than or equal to 0.05, it indicates that the methylation level of this CpG site is unstable between tumor tissue samples;
[0082] d. According to the above method, each CpG site that is significantly different between the healthy and chronic inflammation samples but stable in each sample is marked, and the CpG sites marked as unstable between tumor tissue samples are excluded, and the remaining CpG sites are regarded as CpG sites with stable methylation levels between tumor tissue samples;
[0083] e. For the i-th stable CpG site in the tumor tissue sample, based on its number of methylation reads in the healthy cfDNA sample (denoted as x 1i M ) and the number of unmethylated reads (denoted as ), and the methylation rate p in tumor tissue samples i ', according to the binomial distribution Calculate the probability P(x=x 1i M ); if the obtained probability is less than 0.05, it indicates that there is a significant difference in the methylation level of the CpG site between the healthy cfDNA samples and the tumor tissue samples;
[0084] f. For the i-th stable CpG site in the tumor tissue sample, based on the number of methylation reads in the chronic inflammation cfDNA sample (denoted as x 2i M ) and the number of unmethylated reads (denoted as ), and the methylation rate p in tumor tissue samples i ', according to the binomial distribution Calculate the probability P(x=x 2i M ); if the obtained probability is less than 0.05, it means that there is a significant difference in the methylation level of the CpG site between the chronic inflammation cfDNA samples and the tumor tissue samples;
[0085] g. If the methylation level of the i-th stable CpG site in tumor tissue is significantly different between healthy cfDNA samples and tumor tissue samples, and also significantly different between chronic inflammation cfDNA samples and tumor tissue samples, then the methylation level of the CpG site marked is significantly different between different types of samples;
[0086] h. All CpG sites that are stable in tumor tissue samples are marked with significant differences between different types of samples, and CpG sites whose methylation levels are stable between samples of the same type but significantly different between samples of different types are obtained.
[0087] S2. Estimate the methylation level of CpG sites and the proportion of ctDNA: Divide the DNA fragments in cfDNA into two types: tumor DNA and non-tumor DNA, and construct a statistical model M1 based on CpG sites whose methylation levels are stable between samples of the same type but significantly different between samples of different types:
[0088] P(X,Y|β,θ)=P(X|β,θ)P(Y|β)
[0089]
[0090]
[0091]
[0092]
[0093]
[0094] y i1 , y' i1 : The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m
[0095] y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m
[0096] x i , x' i : The number of methylation reads and total reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m
[0097] β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m
[0098] θ: Detect the proportion of ctDNA in cfDNA samples
[0099] Furthermore, the Expectation-Maximization algorithm (EM) is used to solve the model M1, as follows:
[0100] Initial step (I-step):
[0101] N is the number of sequencing fragments in the cfDNA sample
[0102]
[0103] Expected step (E-step):
[0104]
[0105]
[0106] Maximization step (M-step):
[0107]
[0108]
[0109] If |θ (j+1) -θ (j) |<10 -6 If it holds, the iteration ends.
[0110] S3. Types of cfDNA samples tested by likelihood ratio test: The alternative hypothesis model of likelihood ratio test (detection of cfDNA samples containing ctDNA fragments) is M1 model of S2, while the null hypothesis model (detection of no ctDNA fragments in cfDNA samples) is M0:
[0111] P(X, Y|β)=P(X|β)P(Y|β)
[0112]
[0113]
[0114]
[0115]
[0116] y i1 , y' i1 : The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m
[0117] y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m
[0118] x i , x' i: The number of methylation reads and total reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m
[0119] β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m
[0120] The maximum likelihood estimation of model M0 has an analytical solution, namely The likelihood ratio test statistic based on the M0 model and the M1 model follows a chi-square distribution with a degree of freedom of 1. At a significance level of 0.05 (the corresponding critical value is denoted as ρ), the type of cfDNA sample to be tested is determined as follows:
[0121] (1) Based on healthy cfDNA samples, the likelihood ratio test statistic Δl=2(l1-l0) of the M0 model and the M1 model was calculated, where l0 and l1 are the natural log-likelihood values of the M0 model and the M1 model;
[0122] (2) Calculation of the likelihood ratio test statistic for the M0 model and the M1 model based on chronic inflammation cfDNA samples in and is the natural log-likelihood of the M0 model and the M1 model;
[0123] (3) The cfDNA sample to be tested must be a healthy cfDNA sample and must meet one of the following conditions: or
[0124] (4) The cfDNA sample to be tested is a chronic inflammatory cfDNA sample that meets one of the following conditions: or
[0125] (5) The cfDNA sample tested contains ctDNA fragments and must meet the following conditions:
[0126] S4. Prediction of ctDNA fragments: Based on the likelihood ratio test results of S3, if the original hypothesis is not true, the DNA fragments from tumor tissue in the cfDNA sample are predicted as follows based on the methylation level estimated by S2 and the proportion of ctDNA in the cfDNA sample:
[0127] (1) Calculate the probability that fragment c in the cfDNA sample comes from tumor tissue. The formula is as follows:
[0128]
[0129]
[0130] β i1 , β i2 : Estimated value of methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, …, m
[0131] θ: Estimated value of the proportion of ctDNA in the cfDNA sample
[0132] (2) Based on the probability that each DNA fragment in the cfDNA sample comes from tumor tissue, a random number generated by a pseudo-random number generator is used to determine whether the fragment comes from tumor tissue. For example, the probability that DNA fragment c in the cfDNA sample comes from tumor tissue is p c , the random number generated by the pseudo-random number generator is r c , if r c ≤p c , then fragment c is predicted to be ctDNA.
[0133] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for predicting circulating tumor DNA based on DNA methylation characteristics, characterized in that: The steps include: 1) Select CpG sites whose DNA methylation levels are stable among samples of the same type but significantly different among samples of different types; 2) Construct a statistical model and use the expectation-maximization algprithm (EM) algorithm to estimate the methylation levels of the selected CpG sites in different types of samples and the proportion of ctDNA in the detected cfDNA samples; 3) Use the likelihood ratio test (LRT) to determine the sample type of the cfDNA sample tested; 4) Predict and detect ctDNA fragments in cfDNA samples.
2. The prediction method according to claim 1, characterized in that: The selection in step 1) is as follows: based on healthy cfDNA samples, chronic inflammation cfDNA samples and tumor tissue samples, a binomial distribution is constructed to select CpG sites whose DNA methylation levels are stable between samples of the same type but significantly different between samples of different types.
3. The prediction method according to claim 1-2, characterized in that: In step 2), the DNA fragments in cfDNA are divided into two types: tumor DNA and non-tumor DNA, and the statistical model M1 constructed based on the CpG sites selected in step 1) is: P(X,Y|β,θ)=P(X|β,θ)P(Y|β) y i1 , y' i1 : The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m x i , x' i : The number of methylation reads and total number of reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m θ: Detect the proportion of ctDNA in cfDNA samples.
4. The prediction method according to claims 1-3, characterized in that: The alternative hypothesis model of the likelihood ratio test in step 3) is the M1 model in step 2), and the null hypothesis model M0 is: P(X, Y|β)=P(X|β)P(Y|β) y i1 , y' i1 : The number of methylation reads and total reads at the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples), i = 1, 2, ..., m y i2 , y' i2 : The number of methylation reads and total number of reads at the i-th CpG site in the tumor tissue sample, i = 1, 2, ..., m x i , x' i : The number of methylation reads and total number of reads at the i-th CpG site in the cfDNA sample, i = 1, 2, ..., m β i1 , β i2 : The methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m.
5. The prediction method according to claims 1-4, characterized in that: In step 4), based on the estimated methylation level in step 2) and the proportion of ctDNA in the detected cfDNA sample, as well as the likelihood ratio test result in step 3), the probability that the fragment c in the detected cfDNA sample comes from the tumor tissue is calculated according to the following formula: β i1 , β i2 : Estimated value of methylation level of the i-th CpG site in healthy cfDNA samples (or chronic inflammation cfDNA samples) and tumor tissues, i = 1, 2, ..., m θ: Estimated value of the proportion of ctDNA in the detected cfDNA sample.
6. The evaluation method according to claims 1-5, characterized in that: Based on the probability that fragment c in the cfDNA sample comes from tumor tissue, a random number generated by a pseudo-random number generator is used to determine whether fragment c comes from tumor tissue.