DMR set recognition result evaluation method, evaluation system and selection method

Abstract

The present invention discloses a method for evaluating the results of DMR set identification, including clustering CpG sites with the probe CpG sites in the region as the center for each predicted differentially methylated region; calculating the average methylation level difference of each probe on the methylation chip data of the experimental group and the control group; calculating the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class; calculating the methylation level difference of each differentially methylated region on the methylation chip data of the experimental group and the control group; calculating the evaluation index of each method to be evaluated to complete the evaluation of the DMR set identification result. The present invention also discloses an evaluation system for implementing the DMR set identification result evaluation method, and a selection method including the DMR set identification result evaluation method. The present invention does not need to match additional biological data, can provide guidance for selecting a more excellent differentially methylated region set, and is highly objective, reliable, and scientifically reasonable.

Description

Technical Field

[0001] The invention belongs to the field of computer technology, and in particular relates to a DMR set recognition result evaluation method, an evaluation system and a selection method. Background Art

[0002] In recent years, a large amount of 450K methylation chip data has been generated to study disease-specific methylation changes in the occurrence and development of diseases, as well as tissue-specific methylation phenomena in different tissues. In order to analyze differentially methylated regions (DMRs) between disease tissues and normal tissues and DMRs between different tissues, researchers have developed many methods for DMR identification and used them for the analysis of methylation chip data. According to the identification process, these methods can be divided into methods based on the merging of differentially methylated sites and methods based on the calculation of regional difference significance. Among them, the types of regions can be divided into predefined interval types and custom interval types: predefined interval types usually include functional regions on the genome, such as promoter regions, CpG (Cytosine-phosphoric acid-Guanine, cytosine-phosphate-guanine) islands, and the first 2000bp region of the transcription start site. These regions are usually enriched with CpG sites to divide the intervals; custom regions are mainly based on the genomic coordinate distance corresponding to the probe CpG site on the methylation chip data, and the probe CpG sites with a distance less than a given threshold are divided into an interval.

[0003] However, different methods have different strategies at each step of the DMR identification process, and these steps have an important impact on the final identified DMR set and the methylation differences of DMRs. Faced with numerous DMR identification methods, how to choose a more effective method for DMR identification, or to select a better DMR set from the identification results of multiple methods for downstream research, is particularly important for researchers such as medical scientists and biologists.

[0004] Due to the lack of a standard DMR set, researchers now basically use simulated data to evaluate the accuracy of identified DMRs. However, the hypothetical process of generating simulated data will also introduce the bias of the algorithm designer, so it cannot objectively reflect the real data; at the same time, the enrichment analysis of other biological data, such as chip-seq data of histones and transcription factor binding sites, gene expression data, genomic region annotation information, etc., although it can also be used to support the biological significance of the predicted DMR set, when facing the methylation chip data analysis of different tissues, it is necessary to match chip-seq data, gene expression data and other data; these data that need to be matched have brought great difficulties to different studies and analyses. Summary of the invention

[0005] One of the purposes of the present invention is to provide a DMR set recognition result evaluation method that is highly objective, reliable and scientifically reasonable.

[0006] A second objective of the present invention is to provide an evaluation system for implementing the DMR set recognition result evaluation method.

[0007] A third object of the present invention is to provide a selection method including the DMR set recognition result evaluation method.

[0008] The DMR set recognition result evaluation method provided by the present invention comprises the following steps:

[0009] S1. For each predicted differentially methylated region, cluster all CpG sites in the region with the probe CpG site in the region as the center;

[0010] S2. Calculate the difference in the average methylation level of each probe between the methylation chip data of the experimental group and the methylation chip data of the control group;

[0011] S3. Based on the methylation sequencing data of other tissues in the public database, calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class;

[0012] S4. Based on the average methylation level difference obtained in step S2 and the correlation coefficient obtained in step S3, the methylation level difference between each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group is calculated;

[0013] S5. For each set of differentially methylated regions predicted by the method to be evaluated, calculate the evaluation index of each method to be evaluated, and complete the evaluation of the DMR set identification results.

[0014] The step S1 is specifically to perform K-means clustering on all CpG sites in each predicted differentially methylated region, with the probe CpG site in the region as the center; during clustering, the classification function of the CpG sites in the differentially methylated region based on K-means clustering is calculated and solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the CpG site set represented by each probe CpG site.

[0015] The step S1 specifically includes the following steps:

[0016] S1.1. The following formula is used as the classification function of CpG sites in differentially methylated regions based on K-means clustering:

[0017]

[0018] In the formula For taking and The function of operating on the category number corresponding to the nearest center; CpG site i The genomic coordinates of CpG site for probe j The corresponding genomic coordinates; is the 2-norm; in K-means clustering, k The value is the number of CpG sites of the methylation chip probe contained in the region, and k The center value corresponds to the genomic coordinate of the probe CpG site;

[0019] S1.2. For each differentially methylated region, all CpG sites in the region are clustered using K-means with the probe CpG site in the region as the center. During clustering, the classification function in step S1.1 is solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the set of CpG sites represented by each probe CpG site.

[0020] The step S2 specifically calculates the difference in average methylation level of each probe between the experimental group methylation chip data and the control group methylation chip data based on the methylation level of the CpG site in the sample.

[0021] The step S2 specifically includes the following steps:

[0022] The following formula is used to calculate each probe i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0023]

[0024] In the formula CpG site i In the sample k methylation levels in; CpG site j In the sample k methylation levels in; n A is the number of methylation chip data of the experimental group in the dataset; n B is the number of methylation chip data of the control group in the dataset.

[0025] The step S3 is specifically to calculate the Pearson correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data of other tissues in the public database and the methylation levels of the probe CpG sites and adjacent CpG sites in each sample.

[0026] The Pearson correlation coefficient of each probe CpG site and other CpG sites in the corresponding class is calculated by using the following formula to obtain the Pearson correlation coefficient of two CpG sites: i and j Pearson correlation coefficient between :

[0027]

[0028] In the formula CpG site i In the sample k methylation levels in; CpG site j In the sample k methylation levels in; CpG site i Mean methylation level in all samples; CpG site j Mean methylation level in all samples; n is the number of samples in the data set.

[0029] The calculation of the methylation level difference of each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group in step S4 is specifically to calculate the differentially methylated region using the following formula D i Differences in methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0030]

[0031] In the formula D i Include k Methylation chip probe CpG site, centered on the methylation chip probe CpG site, D i The CpG sites are divided into k Each probe CpG site i Corresponding to a class i , and class i Contains m CpG sites; c im The probe CpG site obtained in step S3 iThe corresponding class i Pearson correlation coefficient of the mth CpG site in ; The probe CpG site obtained in step S2 i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data; n for D i The number of CpG sites included.

[0032] The step S5 is specifically to calculate the evaluation index of each method to be evaluated for the set of differentially methylated regions predicted by each method to be evaluated, based on the difference in methylation levels corresponding to the differentially methylated regions, the total number of probes, the length and the number of CpG sites included, to complete the evaluation of the DMR set recognition results.

[0033] The step S5 specifically includes the following steps:

[0034] S5.1. Based on the set of differentially methylated regions predicted by the band evaluation method, the corresponding methylation level differences were divided into N n For each interval, the total number of CpG sites and probes contained in the differentially methylated region set in each interval were counted, and the total length of the differentially methylated region in each interval was calculated; N n is a positive integer;

[0035] S5.2. Calculate the first evaluation index of the method to be evaluated using the following formula: DMRn ( t ):

[0036]

[0037] In the formula w i For the i The weight value of each interval; n i To fall in i The total number of CpG sites contained in the differentially methylated region set of each interval; N i To fall in i The total number of probes contained in the differentially methylated region set of each interval; L i For the i The total length of the differentially methylated regions in the interval;

[0038] S5.3. Calculate the second evaluation index of the method to be evaluated using the following formula: DMRl ( t ):

[0039]

[0040] In the formula N n The number of intervals divided for the methylation level differences;

[0041] S5.4. Evaluate the DMR set recognition results obtained by the method to be evaluated based on the evaluation indicators calculated in steps S5.2 and S5.3.

[0042] The present invention also provides an evaluation system for implementing the DMR set identification result evaluation method, comprising a CpG site clustering module, an average methylation level difference calculation module, a correlation coefficient calculation module, a methylation level difference calculation module and a DMR set identification result evaluation module in sequence; the average methylation level difference calculation module and the correlation coefficient calculation module are connected in parallel; the CpG site clustering module, the parallel average methylation level difference calculation module and the correlation coefficient calculation module, the methylation level difference calculation module and the DMR set identification result evaluation module are connected in series in sequence; the CpG site clustering module is used to cluster all CpG sites in the region based on each predicted differentially methylated region, with the probe CpG site in the region as the center, and upload the clustering results to the average methylation level difference calculation module and the correlation coefficient calculation module at the same time; the average methylation level difference calculation module is used to calculate the difference between each probe in the experimental group methylation chip data and The average methylation level difference on the control group methylation chip data is calculated, and the calculation result is uploaded to the methylation level difference calculation module; the correlation coefficient calculation module is used to calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data between the experimental group methylation chip data and the control group methylation chip data, and the calculation result is uploaded to the methylation level difference calculation module; the methylation level difference calculation module is used to calculate the methylation level difference between each differentially methylated region on the experimental group methylation chip data and the control group methylation chip data based on the received average methylation level difference and correlation coefficient, and upload the calculation result to the DMR set recognition result evaluation module; the DMR set recognition result evaluation module is used to calculate the evaluation index of each method to be evaluated for the set of differentially methylated regions predicted by each method to be evaluated, complete the evaluation of the DMR set recognition result, and output the evaluation result.

[0043] The present invention also provides a selection method including the DMR set recognition result evaluation method, which specifically includes the following steps:

[0044] S6. According to the evaluation results of the DMR set recognition results obtained in step S5, the DMR set recognition result with the best evaluation result and its corresponding evaluation method are selected as the final selection result.

[0045] The DMR set identification result evaluation method, evaluation system and selection method provided by the present invention take into account that the methylation difference calculation method based on the methylation level of the probe CpG site in the differentially methylated region ignores the influence of the correlation between the probe CpG site and the adjacent genomic CpG site on the regional methylation difference calculation, and therefore proposes a method of fusing the correlation between the probe CpG site and the adjacent genomic CpG site to recalculate the methylation difference of the differentially methylated region; at the same time, the present invention also takes into account that on the same methylation chip data set, the differentially methylated region sets identified by different methods have different methylation difference distributions, the number of probe CpG sites covered and the number of genomic CpG sites covered. The number of CpG sites, genome length, etc. have different characteristics. By dividing the methylation difference into intervals, different weights are assigned to the differentially methylated region sets falling in different intervals, and the different methods are further evaluated by calculating the weighted CpG site number and genome length of the differentially methylated regions greater than a given methylation difference threshold identified by each method. Therefore, the present invention does not require matching additional biological data. By sorting different methods according to the calculated indicators, it can provide guidance for selecting a differentially methylated region set containing a larger number of CpG sites and covering a longer genome length under a given methylation difference threshold, and it is highly objective, reliable, and scientifically reasonable. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The figure is a schematic diagram of the method flow of the evaluation method of the present invention.

[0047] Figure 2 Schematic diagram of the system structure of the evaluation system of the present invention.

[0048] Figure 3 The present invention is a schematic diagram of the method flow of the selection method. DETAILED DESCRIPTION

[0049] like Figure 1 The method flow chart of the evaluation method of the present invention is shown as follows: The DMR set recognition result evaluation method provided by the present invention comprises the following steps:

[0050] S1. For each predicted differentially methylated region, cluster all CpG sites in the region with the probe CpG site in the region as the center; specifically, for each predicted differentially methylated region, cluster all CpG sites in the region with the probe CpG site in the region as the center using K-means clustering; during clustering, calculate and solve the classification function of CpG sites in the differentially methylated region based on K-means clustering to obtain the clustering relationship between CpG sites in the differentially methylated region and probe CpG sites, thereby obtaining the set of CpG sites represented by each probe CpG site;

[0051] The specific implementation includes the following steps:

[0052] S1.1. The following formula is used as the classification function of CpG sites in differentially methylated regions based on K-means clustering:

[0053]

[0054] In the formula For taking and The function of operating on the category number corresponding to the nearest center; CpG site i The genomic coordinates of CpG site for probe j The corresponding genomic coordinates; is the 2-norm; in K-means clustering, k The value is the number of CpG sites of the methylation chip probe contained in the region, and k The center value corresponds to the genomic coordinate of the probe CpG site;

[0055] S1.2. For each differentially methylated region, all CpG sites in the region are clustered using K-means with the probe CpG site in the region as the center; during clustering, the classification function in step S1.1 is solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the CpG site set represented by each probe CpG site;

[0056] S2. Calculate the average methylation level difference of each probe in the methylation chip data of the experimental group and the methylation chip data of the control group; specifically, calculate the average methylation level difference of each probe in the methylation chip data of the experimental group and the methylation chip data of the control group based on the methylation level of the CpG site in the sample;

[0057] The specific implementation includes the following steps:

[0058] The following formula is used to calculate each probe i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0059]

[0060] In the formula CpG site i In the sample k methylation levels in; CpG site j In the sample k methylation levels in;n A is the number of methylation chip data of the experimental group in the dataset; n B is the number of methylation chip data of the control group in the data set;

[0061] S3. Based on the methylation sequencing data of other tissues in the public database, calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class; specifically, based on the methylation sequencing data of other tissues in the public database and the methylation levels of the probe CpG site and the adjacent CpG sites in each sample, calculate the Pearson correlation coefficient between each probe CpG site and other CpG sites in the corresponding class;

[0062] When it comes to specific implementation,

[0063] Two CpG sites were calculated using the following formula: i and j Pearson correlation coefficient between :

[0064]

[0065] In the formula CpG site i In the sample k methylation levels in; CpG site j In the sample k methylation levels in; CpG site i Mean methylation level in all samples; CpG site j Mean methylation level in all samples; n is the number of samples in the data set;

[0066] S4. Based on the average methylation level difference obtained in step S2 and the correlation coefficient obtained in step S3, the methylation level difference between each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group is calculated;

[0067] Specifically, the differentially methylated regions were calculated using the following formula: D i Differences in methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0068]

[0069] In the formula D i Includek Methylation chip probe CpG site, centered on the methylation chip probe CpG site, D i The CpG sites are divided into k Each probe CpG site i Corresponding to a class i , and class i Contains m CpG sites; c im The probe CpG site obtained in step S3 i The corresponding class i Pearson correlation coefficient of the mth CpG site in ; The probe CpG site obtained in step S2 i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data; n for D i The number of CpG sites included

[0070] S5. For each set of differentially methylated regions predicted by the method to be evaluated, the evaluation index of each method to be evaluated is calculated to complete the evaluation of the DMR set recognition result; specifically, for each set of differentially methylated regions predicted by the method to be evaluated, based on the methylation level difference corresponding to the differentially methylated region, the total number of probes, the length and the number of CpG sites included, the evaluation index of each method to be evaluated is calculated to complete the evaluation of the DMR set recognition result;

[0071] The specific implementation includes the following steps:

[0072] S5.1. Based on the set of differentially methylated regions predicted by the band evaluation method, the corresponding methylation level differences were divided into N n intervals (preferably 10, specifically , , …, ), respectively count the total number of CpG sites and the total number of probes contained in the differentially methylated region set falling in each interval, and calculate the total length of the differentially methylated region in each interval; N n is a positive integer;

[0073] S5.2. Calculate the first evaluation index of the method to be evaluated using the following formula: DMRn ( t ):

[0074]

[0075] In the formula wi For the i The weight value of each interval; n i To fall in i The total number of CpG sites contained in the differentially methylated region set of each interval; N i To fall in i The total number of probes contained in the differentially methylated region set of each interval; L i For the i The total length of the differentially methylated regions in the interval;

[0076] S5.3. Calculate the second evaluation index of the method to be evaluated using the following formula: DMRl ( t ):

[0077]

[0078] In the formula N n The number of intervals divided for the methylation level differences;

[0079] S5.4. Evaluate the DMR set recognition results obtained by the method to be evaluated based on the evaluation indicators calculated in steps S5.2 and S5.3.

[0080] The evaluation method of the present invention is further described below in conjunction with an embodiment:

[0081] Bumphunter, ProbeLasso, DMRcate, mCSEA, Combp, Ipdmr, seqlm and other methods were used to predict differentially methylated regions of methylation chip data of 31 prostate cancer samples and 16 normal samples contained in the GSE112047 dataset;

[0082] For each differentially methylated region predicted by each method, it is assumed that there are k probe CpG sites, with their corresponding genomic coordinates as k The center value is used to perform K-means clustering on all CpG sites in the genome in the region. The classification function of CpG sites in differentially methylated regions based on K-means clustering is:

[0083]

[0084] in, CpG site i The genomic coordinates of CpG site for probe jThe corresponding genomic coordinates. i Solve the classification function and classify it into the probe CpG site corresponding to the minimum distance j In the cluster where the CpG sites in the region are located, the clustering relationship between the CpG sites in the region and the probe CpG sites is obtained, and then the CpG site set represented by each probe CpG site is obtained;

[0085] The average methylation level difference of each probe in the prostate cancer sample group and the normal control group in the GSE112047 dataset was calculated. i The formula for calculating the difference in the mean methylation levels between the two groups is as follows:

[0086]

[0087] in, and Represents CpG sites i and j In the sample k The methylation level in n A and n B They represent the number of prostate cancer samples in group A and the number of normal control samples in group B in the dataset respectively;

[0088] The methylation sequencing dataset GSE121721 containing 60 tumor sample data and 4 normal samples was selected. Based on the methylation levels of the probe CpG site and the adjacent CpG sites in each sample, the Pearson correlation coefficient between each probe CpG site and other CpG sites in the class was calculated. i and j The calculation formula of the Pearson correlation coefficient is as follows:

[0089]

[0090] in, and Represents CpG sites i and j The average methylation level in all samples, and Represents CpG sites i and j In the sample k The methylation level in n Represents the number of samples in the data set;

[0091] For each differentially methylated region D iBased on the average methylation level difference of each probe in the two groups of samples and the Pearson correlation coefficient with other CpG sites in the class, the methylation level difference of the region in the methylation chip data of the experimental group and the control group was calculated according to the following formula:

[0092]

[0093] in, D i Included k Methylation chip probe CpG site, centered on the methylation chip probe CpG site, D i The CpG sites are divided into k classes. Each probe CpG site i Corresponding to a class i , assuming class i Contains m CpG sites, c im is the calculated probe CpG site i With class i The Pearson correlation coefficient of the mth CpG site in is the calculated probe CpG site i The difference in average methylation levels between the experimental and control groups in the methylation array data; n for D i The total number of CpG sites included; in methylation sequencing data, there are some CpG sites that are not covered by sequencing fragments or do not match the probe CpG site. i There are valid values ​​in pairs in the sample, so n Set to D i The total number of CpG sites included in that can calculate the Pearson correlation coefficient with the corresponding probe CpG site;

[0094] The above treatment was performed on each differentially methylated region to obtain its methylation level difference;

[0095] For each set of differentially methylated regions predicted by each method, the index was calculated based on the difference in methylation levels, the number of probes, the length, and the number of CpG sites contained in the differentially methylated regions. DMRn and DMR l ;

[0096] The range of methylation level differences [0,1] is divided into 10 intervals, specifically labeled as the set S = {(0,0.1), (0.1,0.2), ..., (0.9,1)}, where each element ( left i , right i) corresponds to an interval, and the total number of CpG sites, the total number of probes, and the total length of the differentially methylated regions in the differentially methylated region set falling in each interval are counted respectively;

[0097] For each set of differentially methylated regions predicted by each method, we calculated t= Indicators at 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8 and 0.9 DMRn ( t )and DMRl (t). Indicators DMRn ( t ) is calculated as follows:

[0098]

[0099]

[0100] in, w i For the i The weight corresponding to the interval is preferably set to i The midpoint value of the interval; n i , N i and L i The methylation level differences were i The total number of CpG sites, the total number of probes, and the total length of the differentially methylated regions included in the differentially methylated region set of each interval;

[0101] Finally, t By comparing the values ​​of the differentially methylated region sets predicted by different methods DMRn ( t )and DMRl ( t ) values ​​and sort them: The set of differentially methylated regions predicted by method A DMRn ( t ) is larger, indicating that the methylation difference predicted by method A is greater than t The differentially methylated region set predicted by method A covers more CpG sites that are correlated with the methylation levels of the probe CpG sites; DMRl ( t ) is larger, indicating that the methylation difference predicted by method A is greater than t The differentially methylated region set covers a longer genomic region that is correlated with the methylation level of the probe CpG site; therefore, the evaluation principle is that if the final DMRn ( t ) value is larger orDMRl ( t ) value, the larger the differentially methylated region set predicted by method A is.

[0102] like Figure 2 The system structure diagram of the prediction system of the present invention is shown as follows: the evaluation system for implementing the DMR set identification result evaluation method provided by the present invention comprises a CpG site clustering module, an average methylation level difference calculation module, a correlation coefficient calculation module, a methylation level difference calculation module and a DMR set identification result evaluation module in sequence; the average methylation level difference calculation module and the correlation coefficient calculation module are connected in parallel; the CpG site clustering module, the parallel average methylation level difference calculation module and the correlation coefficient calculation module, the methylation level difference calculation module and the DMR set identification result evaluation module are connected in series in sequence; the CpG site clustering module is used to cluster all CpG sites in the region based on each predicted differentially methylated region, with the probe CpG site in the region as the center, and upload the clustering results to the average methylation level difference calculation module and the correlation coefficient calculation module at the same time; the average methylation level difference calculation module is used to calculate each The average methylation level difference of each probe on the methylation chip data of the experimental group and the methylation chip data of the control group is calculated, and the calculation result is uploaded to the methylation level difference calculation module; the correlation coefficient calculation module is used to calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data of other tissues in the public database, and the calculation result is uploaded to the methylation level difference calculation module; the methylation level difference calculation module is used to calculate the methylation level difference of each differentially methylated region on the methylation chip data of the experimental group and the methylation chip data of the control group based on the received average methylation level difference and correlation coefficient, and the calculation result is uploaded to the DMR set recognition result evaluation module; the DMR set recognition result evaluation module is used to calculate the evaluation index of each method to be evaluated for the set of differentially methylated regions predicted by each method to be evaluated, complete the evaluation of the DMR set recognition result, and output the evaluation result.

[0103] like Figure 3 The method flow chart of the selection method of the present invention is shown as follows: The selection method provided by the present invention, which includes the DMR set recognition result evaluation method, specifically includes the following steps:

[0104] S1. For each predicted differentially methylated region, cluster all CpG sites in the region with the probe CpG site in the region as the center;

[0105] S2. Calculate the difference in the average methylation level of each probe between the methylation chip data of the experimental group and the methylation chip data of the control group;

[0106] S3. Based on the methylation sequencing data of other tissues in the public database, calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class;

[0107] S4. Based on the average methylation level difference obtained in step S2 and the correlation coefficient obtained in step S3, the methylation level difference between each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group is calculated;

[0108] S5. For each set of differentially methylated regions predicted by the method to be evaluated, the evaluation index of each method to be evaluated is calculated to complete the evaluation of the DMR set identification results;

[0109] S6. According to the evaluation results of the DMR set recognition results obtained in step S5, the DMR set recognition result with the best evaluation result and its corresponding evaluation method are selected as the final selection result.

[0110] The step S1 is specifically to perform K-means clustering on all CpG sites in each predicted differentially methylated region, with the probe CpG site in the region as the center; during clustering, the classification function of the CpG sites in the differentially methylated region based on K-means clustering is calculated and solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the CpG site set represented by each probe CpG site.

[0111] The step S1 specifically includes the following steps:

[0112] S1.1. The following formula is used as the classification function of CpG sites in differentially methylated regions based on K-means clustering:

[0113]

[0114] In the formula For taking and The function of operating on the category number corresponding to the nearest center; CpG site i The genomic coordinates of CpG site for probe j The corresponding genomic coordinates; is the 2-norm; in K-means clustering, k The value is the number of CpG sites of the methylation chip probe contained in the region, and k The center value corresponds to the genomic coordinate of the probe CpG site;

[0115] S1.2. For each differentially methylated region, all CpG sites in the region are clustered using K-means with the probe CpG site in the region as the center. During clustering, the classification function in step S1.1 is solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the set of CpG sites represented by each probe CpG site.

[0116] The step S2 specifically calculates the difference in average methylation level of each probe between the experimental group methylation chip data and the control group methylation chip data based on the methylation level of the CpG site in the sample.

[0117] The step S2 specifically includes the following steps:

[0118] The following formula is used to calculate each probe i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0119]

[0120] In the formula CpG site i In the sample k methylation levels in; CpG site j In the sample k methylation levels in; n A is the number of methylation chip data of the experimental group in the dataset; n B is the number of methylation chip data of the control group in the dataset.

[0121] The step S3 is specifically to calculate the Pearson correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data of other tissues in the public database and the methylation levels of the probe CpG sites and adjacent CpG sites in each sample.

[0122] The Pearson correlation coefficient of each probe CpG site and other CpG sites in the corresponding class is calculated by using the following formula to obtain the Pearson correlation coefficient of two CpG sites: i and j Pearson correlation coefficient between :

[0123]

[0124] In the formula CpG site i In the sample kmethylation levels in; CpG site j In the sample k methylation levels in; CpG site i Mean methylation level in all samples; CpG site j Mean methylation level in all samples; n is the number of samples in the data set.

[0125] The calculation of the methylation level difference of each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group in step S4 is specifically to calculate the differentially methylated region using the following formula D i Differences in methylation levels between the experimental group methylation chip data and the control group methylation chip data :

[0126]

[0127] In the formula D i Include k Methylation chip probe CpG site, centered on the methylation chip probe CpG site, D i The CpG sites are divided into k Each probe CpG site i Corresponding to a class i , and class i Contains m CpG sites; c im The probe CpG site obtained in step S3 i The corresponding class i Pearson correlation coefficient of the mth CpG site in ; The probe CpG site obtained in step S2 i The difference in average methylation levels between the experimental group methylation chip data and the control group methylation chip data; n for D i The number of CpG sites included.

[0128] The step S5 is specifically to calculate the evaluation index of each method to be evaluated for the set of differentially methylated regions predicted by each method to be evaluated, based on the difference in methylation levels corresponding to the differentially methylated regions, the total number of probes, the length and the number of CpG sites included, to complete the evaluation of the DMR set recognition results.

[0129] The step S5 specifically includes the following steps:

[0130] S5.1. Based on the set of differentially methylated regions predicted by the band evaluation method, the corresponding methylation level differences were divided into N n For each interval, the total number of CpG sites and probes contained in the differentially methylated region set in each interval were counted, and the total length of the differentially methylated region in each interval was calculated; N n is a positive integer;

[0131] S5.2. Calculate the first evaluation index of the method to be evaluated using the following formula: DMRn ( t ):

[0132]

[0133] In the formula w i For the i The weight value of each interval; n i To fall in i The total number of CpG sites contained in the differentially methylated region set of each interval; N i To fall in i The total number of probes contained in the differentially methylated region set of each interval; L i For the i The total length of the differentially methylated regions in the interval;

[0134] S5.3. Calculate the second evaluation index of the method to be evaluated using the following formula: DMRl ( t ):

[0135]

[0136] In the formula N n The number of intervals divided for the methylation level differences;

[0137] S5.4. Evaluate the DMR set recognition results obtained by the method to be evaluated based on the evaluation indicators calculated in steps S5.2 and S5.3.

[0138] The step S6 is specifically the evaluation index DMRn ( t )and DMRl ( t) is larger, the better the evaluation result is; the DMR set identification result with the best evaluation result and its corresponding evaluation method are selected as the final selection result, so as to provide or select a better DMR set for researchers for subsequent research.

[0139] The present invention clusters the probe CpG sites with other CpG sites in the differentially methylated region, based on the correlation between the probe CpG sites and the represented CpG sites and the average methylation difference of the probe CpG sites in two groups of methylation chip data, so as to more accurately estimate the methylation difference of each differentially methylated region; DMRn ( t )and DMRl ( t ) By dividing the methylation difference into intervals, different weights positively correlated with the methylation difference are assigned to different intervals, and the number of CpG sites covered by the differentially methylated regions predicted by the weighted calculation method, the number of CpG sites and the length of the probe CpG sites are greater than the given threshold, and the differentially methylated region sets predicted by different methods are evaluated based on this information. Based on the above two advantages, the accuracy of the methylation difference calculation of the differentially methylated regions is improved, and the differentially methylated region sets are evaluated by integrating the information such as the number of CpG sites covered by the differentially methylated regions, the number of CpG sites and the length of the probe, so as to achieve an effective evaluation of the differentially methylated region identification method on the methylation chip.

Claims

1. A DMR set recognition result evaluation method, characterized in that The steps include: S1. For each predicted differentially methylated region, cluster all CpG sites in the region with the probe CpG site in the region as the center; S2. Calculate the difference in the average methylation level of each probe between the experimental group methylation chip data and the control group methylation chip data; S3. Based on the methylation sequencing data of other tissues in the public database, calculate the correlation coefficient of each probe CpG site with other CpG sites in the corresponding class; S4. Based on the average methylation level difference obtained in step S2 and the correlation coefficient obtained in step S3, the methylation level difference between each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group is calculated; S5. For each set of differentially methylated regions predicted by the method to be evaluated, the evaluation index of each method to be evaluated is calculated to complete the evaluation of the DMR set identification results; Specifically, for each set of differentially methylated regions predicted by the method to be evaluated, based on the methylation level differences, the total number of probes, the length, and the number of CpG sites included in the differentially methylated regions, the evaluation index of each method to be evaluated is calculated to complete the evaluation of the DMR set identification results; The specific implementation includes the following steps: S5.

1. According to the set of differentially methylated regions predicted by the method to be evaluated, the corresponding methylation level differences are divided into N n intervals, and count the total number of CpG sites and probes contained in the differentially methylated region set in each interval, and calculate the total length of the differentially methylated region in each interval; N n is a positive integer; S5.

2. Calculate the first evaluation index DMRn(t) of the method to be evaluated using the following formula: Where w i is the weight value of the ith interval; n i N is the total number of CpG sites contained in the set of differentially methylated regions falling in the i-th interval; i is the total number of probes contained in the differentially methylated region set falling in the i-th interval; L i is the total length of the differentially methylated region in the ith interval; S5.

3. Calculate the second evaluation index DMRl(t) of the method to be evaluated using the following formula: Where N n The number of intervals divided for the methylation level differences; S5.

4. Evaluate the DMR set recognition results obtained by the method to be evaluated based on the evaluation indicators calculated in steps S5.2 and S5.

3.

2. The DMR set recognition result evaluation method according to claim 1, characterized in that The step S1 is specifically to perform K-means clustering on all CpG sites in each predicted differentially methylated region, with the probe CpG site in the region as the center; during clustering, the classification function of the CpG sites in the differentially methylated region based on K-means clustering is calculated and solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the CpG site set represented by each probe CpG site.

3. The DMR set recognition result evaluation method according to claim 2, characterized in that The step S1 specifically includes the following steps: S1.

1. The following formula is used as the classification function of CpG sites in differentially methylated regions based on K-means clustering: In the formula To take and X (i) The function of the category number operation corresponding to the nearest center; X (i) is the genomic coordinate of CpG site i; μ j is the genomic coordinate corresponding to the probe CpG site j; || || is the 2-norm; in K-means clustering, k is the number of methylation chip probe CpG sites contained in the region, and the k center values ​​correspond to the genomic coordinates of the probe CpG site; S1.

2. For each differentially methylated region, all CpG sites in the region are clustered using K-means with the probe CpG site in the region as the center. During clustering, the classification function in step S1.1 is solved to obtain the clustering relationship between the CpG sites in the differentially methylated region and the probe CpG sites, thereby obtaining the set of CpG sites represented by each probe CpG site.

4. The DMR set recognition result evaluation method according to claim 2, characterized in that The step S2 is specifically to calculate the average methylation level difference between the experimental group methylation chip data and the control group methylation chip data for each probe based on the methylation level of the CpG site in the sample; the average methylation level difference Δβ between the experimental group methylation chip data and the control group methylation chip data for each probe i is calculated using the following formula: i : Where β i,k is the methylation level of CpG site i in sample k; β j,k is the methylation level of CpG site j in sample k; n A is the number of methylation chip data of the experimental group in the dataset; n B is the number of methylation chip data of the control group in the dataset.

5. The DMR set recognition result evaluation method according to claim 4, characterized in that The step S3 is specifically to calculate the Pearson correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data of other tissues in the public database and the methylation levels of the probe CpG site and the adjacent CpG sites in each sample; Specifically, the Pearson correlation coefficient p between two CpG sites i and j is calculated using the following formula: i,j : Where β i,k is the methylation level of CpG site i in sample k; β i,k is the methylation level of CpG site j in sample k; is the average methylation level of CpG site i in all samples; is the average methylation level of CpG site j in all samples; n is the number of samples in the data set.

6. The DMR set recognition result evaluation method according to claim 5, characterized in that Step S4 is to calculate the difference in methylation level of each differentially methylated region in the methylation chip data of the experimental group and the methylation chip data of the control group, specifically, to calculate the differentially methylated region D using the following formula: i The difference in methylation levels between the experimental group methylation chip data and the control group methylation chip data is dif(D i ): Where D i Contains k methylation chip probe CpG sites, with the methylation chip probe CpG site as the center, D i The CpG sites are divided into k classes, each probe CpG site i corresponds to a class i, and class i contains m CpG sites; c im is the Pearson correlation coefficient between the probe CpG site i obtained in step S3 and the mth CpG site in the corresponding class i; Δβ i is the average methylation level difference between the experimental group methylation chip data and the control group methylation chip data obtained in step S2; n is D i The number of CpG sites included.

7. An evaluation system for implementing the DMR set recognition result evaluation method according to any one of claims 1 to 6, characterized in that It includes a CpG site clustering module, an average methylation level difference calculation module, a correlation coefficient calculation module, a methylation level difference calculation module and a DMR set identification result evaluation module in sequence; the average methylation level difference calculation module and the correlation coefficient calculation module are connected in parallel; the CpG site clustering module, the average methylation level difference calculation module and the correlation coefficient calculation module after the parallel connection, the methylation level difference calculation module and the DMR set identification result evaluation module are connected in series in sequence; the CpG site clustering module is used to cluster all CpG sites in the region based on each predicted differential methylation region, with the probe CpG site in the region as the center, and upload the clustering results to the average methylation level difference calculation module and the correlation coefficient calculation module at the same time; The average methylation level difference calculation module is used to calculate the average methylation level difference of each probe in the methylation chip data of the experimental group and the methylation chip data of the control group, and upload the calculation result to the methylation level difference calculation module; The correlation coefficient calculation module is used to calculate the correlation coefficient between each probe CpG site and other CpG sites in the corresponding class based on the methylation sequencing data of other tissues in the public database, and upload the calculation results to the methylation level difference calculation module; The methylation level difference calculation module is used to calculate the methylation level difference between the experimental group methylation chip data and the control group methylation chip data for each differentially methylated region based on the received average methylation level difference and correlation coefficient, and upload the calculation result to the DMR set identification result evaluation module; the DMR set identification result evaluation module is used to calculate the evaluation index of each method to be evaluated for the set of differentially methylated regions predicted by each method to be evaluated, complete the evaluation of the DMR set identification result, and output the evaluation result.

8. A selection method comprising the DMR set recognition result evaluation method according to any one of claims 1 to 6, characterized in that The following steps are also included: S6. According to the evaluation results of the DMR set recognition results obtained in step S5, the DMR set recognition result with the best evaluation result and its corresponding evaluation method are selected as the final selection result.

Images