Method for detecting and classifying abnormal DNA methylation fragments in cell free DNA

Abstract

The invention belongs to the technical field of bioinformatics, and particularly relates to a method for detecting abnormal DNA methylation fragments from cell free DNA (cell free DNA, cfDNA) and classifying the abnormal DNA methylation fragments from the cell free DNA, and the method comprises the following steps: based on a healthy cfDNA sample and a tumor cfDNA sample, selecting methylation haplotype regions (Methylation Haplotype Block, MHB) having significant difference between the two samples by adopting seven DNA methylation measurement indexes; on the basis of the selected methylated haplotype regions with significant differences, constructing total distribution of seven DNA methylation measurement indexes in the healthy cfDNA sample and the tumor cfDNA sample, and carrying out differential analysis; seven DNA methylation measurement indexes are adopted to construct classification models for abnormal DNA methylation fragments, and the classification models of the seven abnormal DNA methylation fragments are integrated on the basis of differential analysis results of overall distribution of the seven DNA methylation measurement indexes. According to the application, the abnormal DNA methylation fragments in the cell free DNA are detected by adopting seven DNA methylation measurement indexes, and the abnormal DNA methylation fragment classification model based on the seven DNA methylation measurement indexes is integrated, so that the improvement of the sensitivity of tumor early screening is facilitated.

Description

Technical Field

[0001] The present application belongs to the technical field of bioinformatics, and specifically relates to a method for detecting and classifying abnormal DNA methylation fragments from cell free DNA (cfDNA). Technical Background

[0002] 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. Since DNA methylation patterns are cell-specific, there are significant differences in DNA methylation patterns on certain segments of the genome between normal cells and tumor cells. If the above-mentioned genomic segments can be found, DNA fragments from these segments (called abnormal DNA methylation fragments) can be detected from cfDNA, and it can be determined whether these abnormal DNA methylation fragments come from normal cells or tumor cells. Moreover, abnormal DNA methylation fragments already exist in cfDNA in the early stages of cancer, which provides the necessary conditions for the detection of early cancer.

[0003] In order to find the regions in the genome where the DNA methylation patterns differ significantly between normal cells and tumor cells, it is necessary to construct an index that can measure the degree of DNA methylation to distinguish different DNA methylation patterns. Currently, there are seven commonly used DNA methylation metrics, namely: Methylation Entropy (ME), Proportion of Discordant Reads (FDR), Methylated Haplotype Load (MHL), Cell Heterogeneity-Adjusted cLonal Methylation (CHALM), Methylation Concurrence Ratio (MCR), Mean Mehtylation (MM) and Traditional Mehtylation (TM). However, the above DNA methylation metrics have the following two shortcomings: (1) For some DNA methylation patterns with differences, the measurement results are the same; (2) For some DNA methylation patterns with obvious differences, the measurement results are very close. Therefore, the DNA methylation pattern of a certain segment of the genome may not be significantly different under one DNA methylation metric, but it is likely to be significantly different under another DNA methylation metric. Since the DNA fragments in cfDNA are not evenly distributed throughout the genome, detecting abnormal DNA methylation fragments based on only one DNA methylation metric will reduce the sensitivity of the detection.

[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 uses seven DNA methylation metrics to select methylation haplotype blocks (MHB) with significant differences between the two samples based on healthy cfDNA samples and tumor cfDNA samples; based on the selected methylation haplotype regions with significant differences, the overall distribution of the seven DNA methylation metrics in healthy cfDNA samples and tumor cfDNA samples is constructed and differential analysis is performed; the seven DNA methylation metrics are used to construct classification models for abnormal DNA methylation fragments respectively, and based on the differential analysis results of the overall distribution of the seven DNA methylation metrics, the seven abnormal DNA methylation fragment classification models are integrated.

[0006] Therefore, the core purpose of this application is to provide a method for detecting and classifying abnormal DNA methylation fragments in cell-free DNA.

[0007] To achieve the above objectives, this application proposes the following technical solutions:

[0008] 1) Seven DNA methylation metrics were used to select methylation haplotype blocks (MHBs) that were significantly different between healthy cfDNA samples and tumor cfDNA samples;

[0009] 2) construct the overall distribution of seven DNA methylation metrics in healthy cfDNA samples and tumor cfDNA samples and perform differential analysis;

[0010] 3) Seven DNA methylation metrics were used to construct classification models for abnormal DNA methylation fragments;

[0011] 4) Integrate an abnormal DNA methylation fragment classification model based on seven DNA methylation metrics.

[0012] Furthermore, the selection of step 1) is as follows: based on healthy cfDNA samples and tumor cfDNA samples, the values ​​of seven DNA methylation metrics in each methylation haplotype region are calculated, and the values ​​of each DNA methylation metric between the two samples are differentially analyzed, and then the methylation haplotype regions with significant differences are selected.

[0013] Furthermore, the selection is specifically as follows: a. Based on healthy cfDNA samples, the genome is divided into several methylation haplotype regions using the principle of linkage disequilibrium, so that each methylation haplotype region contains several methylation sites and is linked, while the methylation sites between different methylation haplotype regions are independent of each other; b. Seven commonly used DNA methylation metrics are selected, namely: Methylation Entropy (ME), Proportion of Discordant Reads (FDR), Methylated Haplotype Load (MHL), Cell Heterogeneity-Adjusted cLonal Methylation (CHALM), Methylation Concurrence Ratio (MCR), Mean Mehtylation (MM) and Traditional Methylation (Traditional Mehtylation, TM); c. Based on healthy cfDNA samples and tumor cfDNA samples, calculate the values ​​of seven DNA methylation metrics in each methylation haplotype region, and perform differential analysis on the values ​​of each DNA methylation metric between the two samples; d. For a certain methylation haplotype region, if there is at least one methylation metric with a significant difference in value between the two samples, then the region is a methylation haplotype region with significant difference.

[0014] Furthermore, the construction of step 2) is as follows: based on the significantly different methylation haplotype regions selected in step 1), the values ​​of seven DNA methylation metrics in each methylation haplotype region of healthy cfDNA samples and tumor cfDNA samples are calculated, and the frequency distribution of each DNA methylation metric between the two samples is constructed based on the values, and a chi-square test is performed to determine whether the constructed frequency distribution is significantly different between the two samples.

[0015] Furthermore, the construction of step 3) is as follows: based on healthy cfDNA samples and tumor cfDNA samples, the values ​​of the seven DNA methylation metrics on the significantly different methylation haplotype regions selected in step 1) are used as features, and a classic supervised learning algorithm is selected to construct a classification model for the abnormal DNA methylation fragments.

[0016] Furthermore, the integration of step 4) is as follows: based on the chi-square test statistic of each DNA methylation metric in step 2), weighting the corresponding classification model constructed in step 3) is performed, and the weighted classification models are summed.

[0017] Furthermore, the weighting formula is:

[0018]

[0019] w i : The weight of the abnormal DNA methylation fragment classification model based on the i-th DNA methylation metric, i = 1, 2, ..., 7

[0020] χ 2 i : The chi-square test statistic value of the i-th DNA methylation metric, i = 1, 2, ..., 7

[0021] Beneficial technical effects of this application:

[0022] This application selects seven commonly used DNA methylation metrics to identify methylation haplotype regions with significant differences between healthy cfDNA samples and tumor cfDNA samples, and constructs a classification model for DNA fragments from methylation haplotype regions with significant differences based on these seven DNA methylation metrics, and integrates the seven classification models by weight, so that the final classification model takes into account the advantages of each DNA methylation metric according to the weight, providing necessary technical support for improving the sensitivity of early cancer screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 , the present application is a flow chart for the detection and classification of abnormal DNA methylation fragments in cell-free DNA. DETAILED DESCRIPTION

[0024] 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.

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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.

[0029] The present application is described below in conjunction with specific embodiments.

[0030] Example Detection and Classification Method Establishment

[0031] like Figure 1 As shown, the present application provides a method for detecting and classifying abnormal DNA methylation fragments in cell-free DNA, comprising the following steps:

[0032] S1. Selecting Methylation Haplotype Block (MHB) with significant differences: Based on healthy cfDNA samples and tumor cfDNA samples, calculate the values ​​of seven DNA methylation metrics in each methylation haplotype block, and perform differential analysis on the values ​​of each DNA methylation metric between the two samples, and then select the methylation haplotype blocks with significant differences. The specific operations are as follows:

[0033] (1) Based on healthy cfDNA samples, the genome is divided into several methylation haplotype regions using the linkage disequilibrium principle, so that each methylation haplotype region contains several methylation sites and is linked, while the methylation sites between different methylation haplotype regions are independent of each other, including the following three steps:

[0034] a. Divide the methylation haplotype region: Taking a given CpG site as the center, expand 100 bp upstream and downstream respectively until the number of CpG sites is not less than 3 and the expanded 100 bp does not contain any CpG sites;

[0035] b. Based on the methylation haplotype regions divided in step a), linkage disequilibrium analysis is performed on the adjacent CpG sites between any two adjacent methylation haplotype regions. If the linkage disequilibrium coefficient r 2If it is greater than 0.5, it will be merged into one methylation haplotype region;

[0036] c. Based on the methylation haplotype regions merged in step b), linkage disequilibrium analysis is performed on adjacent CpG sites in each methylation haplotype region. If the linkage disequilibrium coefficient r 2 If the value is less than or equal to 0.5 and the number of CpG sites on the left and right sides of the center of the adjacent CpG site is not less than 3, the methylated haplotype region is split into two methylated haplotype regions at the center of the adjacent CpG site until there are no adjacent CpG sites to be divided.

[0037] (2) Seven commonly used DNA methylation metrics were selected, namely: Methylation Entropy (ME), Proportion of Discordant Reads (FDR), Methylated Haplotype Load (MHL), Cell Heterogeneity-Adjusted Clonal Methylation (CHALM), Methylation Concurrence Ratio (MCR), Mean Mehtylation (MM) and Traditional Mehtylation (TM). The definitions of the seven DNA methylation metrics are as follows:

[0038]

[0039] b: Number of CpG sites in a given methylated haplotype region

[0040] n i : The number of sequencing reads with the i-th methylation pattern in a given methylation haplotype region

[0041] N: number of sequencing reads for a given methylation haplotype region

[0042]

[0043] m: The number of non-comethylated (sequencing reads containing both methylated CpG sites and unmethylated CpG sites) sequencing reads in a given methylation haplotype region

[0044] N: number of sequencing reads for a given methylation haplotype region

[0045]

[0046] ι: Maximum number of CpG sites contained in a single sequencing read of a given methylation haplotype region

[0047] p(MH i ): The average percentage of sequencing reads that are methylated at i adjacent CpG sites in a given methylation haplotype region

[0048]

[0049] m: The number of sequencing reads containing methylated CpG sites in a given methylation haplotype region

[0050] N: number of sequencing reads for a given methylation haplotype region

[0051]

[0052] m: The number of unmethylated CpG sites in the sequencing reads containing methylated CpG sites in a given methylated haplotype region

[0053] N: The number of CpG sites in the sequencing reads of a given methylated haplotype region

[0054]

[0055] m: The number of unmethylated CpG sites in the sequencing reads of a given methylated haplotype region

[0056] N: The number of CpG sites in the sequencing reads of a given methylated haplotype region

[0057] Traditional Mehtylation (TM): The average methylation level of CpG sites in a given methylation haplotype region.

[0058] (3) The values ​​of the seven DNA methylation metrics in each methylation haplotype region were calculated, and the Kolmogoroc-Smirnov two-sample test was performed on the values ​​of each DNA methylation metric in the corresponding haplotype region of healthy cfDNA samples and tumor cfDNA samples. If the p-value under a certain DNA methylation metric was less than 0.05, the region was marked as a methylation haplotype region with significant difference.

[0059] S2.1. Constructing a classification model for abnormal DNA methylation fragments: Based on healthy cfDNA samples and tumor cfDNA samples, the values ​​of seven DNA methylation metrics in the selected significantly different methylation haplotype regions are used as features, and the Logistic regression algorithm is selected to construct classification models for abnormal DNA methylation fragments. The specific operations are as follows:

[0060] (1) Establishing a training set: Based on healthy cfDNA samples and tumor cfDNA samples, select DNA fragments from regions with significantly different methylation haplotypes (i.e., abnormal DNA methylation fragments) and annotate them with healthy or tumor labels.

[0061] (2) Selection of classification features: Based on the significantly different methylation haplotype regions and seven DNA methylation metrics (ME, FDR, MHL, CHALM, MCR, MM and TM), the value of each DNA methylation metric in the corresponding methylation haplotype region was calculated and used as the classification feature.

[0062] (3) Training the classification model: Based on the training set and classification features, the Logistic regression algorithm was used to train the regression coefficients of each DNA methylation metric to obtain the corresponding classification model, where the classification boundary value of the Sigmoid function was 0.5.

[0063] S2.2. Construct the overall distribution and differential analysis of DNA methylation metrics: Based on the training set, calculate the values ​​of the seven DNA methylation metrics in each significantly differentially methylated haplotype region, construct the frequency distribution of each DNA methylation metric between healthy cfDNA samples and tumor cfDNA samples based on the values, and perform a chi-square test to determine whether the constructed frequency distribution is significantly different between the two samples. The specific operations are as follows:

[0064] (1) Constructing frequency distribution: Based on the tumor cfDNA samples in the training set, for each DNA methylation metric, the interval [0, 1] was divided into 100 equal parts, and the proportion of methylation haplotype regions with significant differences in values ​​falling in each sub-interval was counted; if the proportion of a sub-interval was 0, the adjacent sub-interval on the right was merged until the proportion was greater than 0, and then k sub-intervals and corresponding proportions p were obtained. i (i=1, 2, ..., k). Based on the healthy cfDNA samples in the training set, for each DNA methylation metric, the number of methylation haplotype regions with significant differences in values ​​falling within the above k subintervals is counted, denoted as f i (i=1, 2,…, k).

[0065] (2) Constructing a chi-square test: For the DNA methylation metric j, a chi-square test is performed on the healthy cfDNA samples and tumor cfDNA samples in the training set, with a degree of freedom of k. j -1, the statistics are as follows:

[0066]

[0067] k j : The number of sub-intervals divided for the DNA methylation metric j

[0068] For the DNA methylation metric j, based on the healthy cfDNA samples in the training set, the number of methylation haplotype regions with significant differences in values ​​falling in the i-th subinterval

[0069] For the DNA methylation metric j, based on the tumor cfDNA samples in the training set, the proportion of methylation haplotype regions with significant differences in values ​​falling in the i-th subinterval

[0070] m: The number of selected significantly different methylation haplotype regions

[0071] n: the number of healthy cfDNA samples in the training set

[0072] S3. Integrated abnormal DNA methylation fragment classification model: The corresponding classification model is weighted based on the value of the chi-square test statistic of each DNA methylation metric, and the weighted classification model is summed. The weighting formula is as follows:

[0073]

[0074] w j : The weight of the abnormal DNA methylation fragment classification model based on the jth DNA methylation metric

[0075] χ 2 j : The chi-square test statistic value of the jth DNA methylation metric

[0076] S4. Detection and classification of abnormal DNA methylation fragments: Given the sequencing data of a cfDNA sample, detect and classify abnormal DNA methylation fragments according to the following steps:

[0077] (1) Comparing the sequencing data of cfDNA samples, the DNA fragments falling in the methylation haplotype regions with significant differences are abnormal DNA methylation fragments.

[0078] (2) Based on the detected abnormal DNA methylation fragments, the value of each DNA methylation metric in the methylation haplotype region with significant differences is calculated and introduced into the corresponding classification model.

[0079] (3) Based on the weight of each classification model, the value of the integrated classification model is calculated and classified, and the classification boundary value is 0.5.

[0080] 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 detecting and classifying abnormal DNA methylation fragments in cell-free DNA, characterized in that: The steps include: 1) Based on healthy cfDNA samples and tumor cfDNA samples, seven DNA methylation metrics were used to select methylation haplotype blocks (MHBs) with significant differences between the two samples; 2) Based on the selected significantly differentially methylated haplotype regions, the overall distribution of seven DNA methylation metrics in healthy cfDNA samples and tumor cfDNA samples was constructed and differentially analyzed; 3) Seven DNA methylation metrics were used to construct classification models for abnormal DNA methylation fragments respectively, and based on the differential analysis results of the overall distribution of the seven DNA methylation metrics, seven abnormal DNA methylation fragment classification models were integrated.

2. The prediction method according to claim 1, characterized in that: The selection of step 1) is as follows: based on healthy cfDNA samples and tumor cfDNA samples, the values ​​of seven DNA methylation metrics are calculated on each methylation haplotype region. If there is a significant difference in the value of a DNA methylation metric between the two samples in a certain methylation haplotype region, then the region is a significantly different methylation haplotype region, and the DNA fragment from the region is an abnormal DNA methylation fragment. The seven DNA methylation metrics are: Methylation Entropy (ME), Proportion of Discordant Reads (FDR), Methylated Haplotype Load (MHL), Cell Heterogeneity-Adjusted cLonal Methylation (CHALM), Methylation Concurrence Ratio (MCR), Mean Mehtylation (MM) and Traditional Mehtylation (TM).

3. The prediction method according to claim 1-2, characterized in that: The step 2) constructs the overall distribution of seven DNA methylation metrics in healthy cfDNA samples and tumor cfDNA samples based on the significantly different methylation haplotype regions selected in step 1), and uses the chi-square test to perform differential analysis.

4. The prediction method according to claims 1-3, characterized in that: In the step 3), the values ​​of the seven DNA methylation metrics on the significantly differentially methylated haplotype regions selected in step 1) are used as features to construct classification models for abnormal DNA methylation fragments respectively, and based on the differential analysis results of step 2), the classification models for abnormal DNA methylation fragments based on the seven DNA methylation metrics are weighted integrated.

Images