DNA methylation and chromatin open data analysis system and method

Abstract

The invention discloses a DNA methylation and chromatin open data analysis system, which comprises a data receiving and processing module, which is used for receiving a DNA sequencing file of a biological sample, carrying out filtering treatment and deduplication treatment, taking methylation site information and generating a site report file; the data statistics and calculation module is used for calculating a methylation ratio and a chromatin openness ratio for a chromosome physical fragment specified by a user according to the site report file; and the data association characterization module is used for carrying out filling processing and dimension reduction analysis on the NA values, and generating a dimension reduction graph for characterizing a clustering relationship between samples and a heat map for displaying methylation and openness relevance. According to the method, DNA methylation and chromatin openness data generation can be integrated, NA filling and MOFA dimension reduction are combined, the collaborative regulation relation between DNA methylation and chromatin openness is disclosed, and an efficient solution is provided for gene expression regulation and disease mechanism research.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and more specifically, to a DNA methylation and chromatin accessibility data analysis system and method. Background Art

[0002] DNA methylation and chromatin accessibility are core indicators of epigenetic regulation, and their joint analysis is of great significance for revealing the regulatory mechanism of gene expression. However, the existing technology has the following problems: Low data integration efficiency: Methylation and open data are usually processed through independent processes, resulting in inconsistent data formats, difficulty in integration, and difficulty in performing correlation analysis under the same framework.

[0003] Subjective NA value processing: In the calculation of methylation-to-openness ratios, NA value filling often relies on simple averaging or direct elimination, lacking biological background considerations (such as functional annotation region priority), and is prone to introducing noise or losing key site information.

[0004] Single dimensionality reduction analysis: Traditional methods (such as PCA) only reduce the dimensionality of a single data type and cannot capture the dynamic synergistic relationship between methylation and openness at the sample level. In addition, the weight distribution is fixed and it is difficult to adapt to data heterogeneity.

[0005] Functional annotation is disconnected from statistical results: After differential region screening, enrichment analysis often relies on manual matching of databases, which is inefficient and prone to missing key pathways, and lacks an automated verification mechanism.

[0006] The root cause of these problems lies in the fragmented design of existing tool chains and the lack of cross-data type correlation analysis methods. Developing an integrated system that implements a closed-loop process from data preprocessing to functional verification is a technical difficulty that urgently needs to be addressed and is the motivation for this invention. Summary of the Invention

[0007] The present invention provides a DNA methylation and chromatin accessibility data analysis system and method, which can integrate DNA methylation and chromatin accessibility data generation, combine NA filling and MOFA dimensionality reduction, reveal the coordinated regulatory relationship between DNA methylation and chromatin accessibility, and provide an efficient solution for gene expression regulation and disease mechanism research.

[0008] In order to achieve these purposes and other advantages according to the present invention, a DNA methylation and chromatin accessibility data analysis system and method are provided, comprising: DNA methylation and chromatin openness data analysis system, including: The data receiving and processing module receives the DNA sequencing file of the biological sample, performs filtering, compares the filtered sequencing file with the reference genome, removes duplicates, extracts methylation site information, sets the parameter --nome-seq, distinguishes CpG methylation sites from GpC open sites, and generates a site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation counts, unmethylation counts, open counts, and non-open counts; The data statistics and calculation module, based on the site report file, selects the methylation site data and open site data falling within the user-specified chromosome physical segment, and calculates the methylation ratio and chromatin openness ratio respectively; The data association characterization module fills the NA values ​​in the methylation ratio and the openness ratio, inputs the filled methylation ratio and openness ratio data into the MOFA model for dimensionality reduction analysis, generates a dimensionality reduction graph representing the clustering relationship between samples and a heat map showing the correlation between methylation and openness, and determines the correlation between methylation and chromatin openness of the specified chromosome physical segment.

[0009] The DNA methylation and chromatin accessibility data analysis method includes the following steps: Filtering the DNA sequencing file of the biological sample to generate a filtered sequencing file; Compare the filtered sequencing file with the reference genome to generate a BAM file; Deduplication is performed on the BAM file to generate a deduplicated BAM file; Extract methylation site information from the deduplicated BAM file and generate a COV file of methylation coverage; Process the COV file and set the parameter --nome-seq to distinguish CpG methylation sites from GpC open sites, and generate a site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation numbers, unmethylation numbers, open numbers, and non-open numbers; Based on the site report file, the methylation number and the openness number are counted. For the specified chromosome physical segment, the methylation site data and the open site data falling within the segment are screened to calculate the methylation ratio and chromatin openness ratio respectively; Fill the NA values ​​in the methylation ratio and openness ratio, perform dimensionality reduction analysis on the filled methylation ratio and openness ratio data, generate a dimensionality reduction graph to characterize the clustering relationship of methylation or openness between samples, and generate a heat map to show the correlation between methylation and openness; Based on the analysis results of the dimensionality reduction graph and heat map, the correlation between the methylation of the specified chromosome physical segment and the chromatin openness was determined.

[0010] Preferably, the NA values ​​in the methylation ratio and the openness ratio are filled, and the filling method is specifically as follows: Set up a dynamic threshold judgment module and select a filling strategy based on the proportion of NA values ​​in each row of the methylation site report file; When the NA value ratio of a row is ≤10%, the sequencing depth and genome functional annotation information of the chromosome fragment corresponding to the row are detected; If the sequencing depth is ≥30X and the chromosome fragment is located in the CpG island or gene promoter annotated region, the weighted sliding mean of the sequencing depth of adjacent sites is used for filling, where the weight is calculated by the ratio of the sequencing depth of the current site to the sum of the sequencing depths of all sites in the sliding window; If the sequencing depth is <30X or located in a non-functional annotation region, the sliding window mean of the adjacent sites is directly used for filling; When the NA value of a row accounts for >10%, query whether the chromosome segment belongs to the pre-annotated enhancer or conserved region; If it belongs to an enhancer or conserved region, the median methylation number of all samples in the same functional category region is used for filling; If the functional annotation cannot be matched, the median global methylation number of the physical segment of the chromosome is used for filling; The filled methylation data were input into the subsequent association analysis module.

[0011] Preferably, the filled methylation ratio and openness ratio data are subjected to dimensionality reduction analysis to generate a dimensionality reduction graph to characterize the clustering relationship of methylation or openness between samples, and a heat map is generated to show the correlation between methylation and openness, specifically: A dynamic weight allocation module is set up in the MOFA model to adjust the weights of the two types of data in dimensionality reduction according to the coefficient of variation of the methylation ratio and the openness ratio; The coefficient of variation of the methylation ratio and the coefficient of variation of the openness ratio were calculated for each sample; If the coefficient of variation of the methylation ratio is higher than the coefficient of variation of the openness ratio, the weight assigned to the methylation data is 1.5 times that of the openness data; If the coefficient of variation of the openness ratio is higher than the coefficient of variation of the methylation ratio, the weight assigned to the openness data is 1.5 times that of the methylation data; Generate a dimensionality reduction graph based on the data after dynamic weight assignment, and mark the weight distribution area in the graph to indicate the main source of cluster differences; The weight distribution results were combined with heat map correlation analysis to screen out chromosome physical segments with significant weight differences as key areas for the coordinated regulation of methylation and openness.

[0012] Preferably, in the dynamic weight allocation module, the weight allocation is dynamically adjusted according to the similarity matrix of the methylation ratio and the openness ratio between the samples; By calculating the Euclidean distance or correlation coefficient between samples, independent similarity matrices for methylation data and openness data were generated; If the clustering tightness of the similarity matrix of methylation data is higher than that of open data, the weight of methylation data is increased; If the clustering tightness of the similarity matrix of the open data is higher than that of the methylation data, the weight of the open data is increased; The adjusted weights are input into the MOFA model for dimensionality reduction, and the areas where high-weight data contribute to the clustering results are marked.

[0013] Preferably, the dimensionality reduction process is divided into two stages: preliminary dimensionality reduction and iterative optimization; In the initial dimensionality reduction stage, the default weights are used to generate the initial dimensionality reduction graph and identify the preliminary clustering groups; In the iterative optimization stage, the weights are redistributed and the dimensionality reduction results are updated according to the discrete degree of methylation and openness data within the cluster group; If the dispersion of methylation data in a group is higher than that of open data, the weight of methylation data in the group will be reduced; If the dispersion of open data is higher than that of methylation data, the weight of open data in this group is reduced; Output the final dimensionality reduction graph and mark the clustering change trajectory before and after weight adjustment.

[0014] Preferably, time series data are integrated into the MOFA model, and the methylation ratio and openness ratio at different time points are input as independent views; Dynamically adjust the weight of data at each time point based on the consistency of change trends between time points; If the methylation at a certain time point is highly consistent with the trend of openness change, the data at that time point will be given a higher weight; If the trend is inconsistent, reduce the weight or mark it as an abnormal time point; Generate dimensionality-reduced trajectory plots across time points to demonstrate the co-evolutionary relationship between methylation and openness over time.

[0015] Preferably, an interactive visualization interface is embedded to allow the user to manually adjust the weight parameters based on the preliminary results of the dimensionality reduction graph; Provide a slider control to dynamically modify the weight ratio of methylation and openness data, and update the dimensionality reduction graph in real time; When the user selects a specific clustering area, the weighted values ​​of the two types of data in the area are automatically calculated and the user is prompted to decide whether to adopt them. The user-corrected weight parameters are fed back to the MOFA model to regenerate the dimensionality reduction graph, and the adjustment history is recorded for retrospective use.

[0016] Preferably, an automated enrichment analysis module is set up to perform functional matching on the physical segments of chromosomes with significant differences through a gene function annotation database, wherein the database includes Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG); Calculate the enrichment significance of the successfully matched functional entries to screen out pathways or biological processes related to the coordinated regulation of methylation and openness; Generate a correlation enrichment map, using chromosome physical segments as nodes and functional annotations as labels, and mark the synchronous change direction of methylation ratio and openness ratio within the segment; Based on the enrichment map, a list of high-confidence co-regulatory regions is output. The regions must meet the following conditions at the same time: 1) Match at least one significantly enriched GO term or KEGG pathway; 2) the absolute value of the Pearson correlation coefficient of the methylation-openness ratio was ≥0.6; 3) The region covers at least one known enhancer or transcription factor binding site; The list of co-regulatory regions was superimposed on the original dimensionality reduction map, and the consistency of their spatial distribution was verified by chromosome overlap heat map.

[0017] Preferably, the adaptive enrichment significance threshold is automatically calculated based on the sample size and sequencing depth of the input data; If the sample size is ≤30 or the average sequencing depth is <20X, the FDR threshold is relaxed to <0.2; If the sample size is ≥100 and the average sequencing depth is ≥50X, the FDR threshold is tightened to <0.01.

[0018] The present invention has at least the following beneficial effects: First, the DNA methylation and chromatin accessibility data analysis system of the present invention integrates DNA methylation and chromatin accessibility data generation through a modular process, realizing full-process analysis from raw sequencing data to biological significance mining. Through dimensionality reduction clustering and heat map visualization, it reveals the coordinated regulatory relationship between DNA methylation and chromatin accessibility, reduces manual intervention, greatly improves processing efficiency, and enhances analysis efficiency and biological discovery capabilities.

[0019] Second, the DNA methylation and chromatin accessibility data analysis method of the present invention distinguishes CpG and GpC sites through parameterization to ensure the accuracy of data generation, combines NA filling and MOFA dimensionality reduction to improve the reliability of correlation analysis, and provides strong support for research on gene expression regulation.

[0020] Third, the DNA methylation and chromatin openness data analysis method of the present invention combines dynamic weight allocation with iterative optimization, and flexibly adjusts according to the degree of data discreteness and similarity, making dimensionality reduction more accurate, effectively mining potential information, revealing the intrinsic structure and biological characteristics of methylation and openness data, and improving clustering stability and analysis depth.

[0021] Fourth, the DNA methylation and chromatin accessibility data analysis method of the present invention integrates time series data and develops an interactive interface, which not only captures dynamic epigenetic changes, but also allows users to manually adjust parameters to verify hypotheses, enhances analysis flexibility and result interpretability, and adapts to complex data scenarios.

[0022] Fifth, the DNA methylation and chromatin accessibility data analysis method of the present invention automates enrichment analysis and adaptive threshold adjustment, quickly locates functional regions, and uses triple condition screening to ensure the biological credibility of the results. At the same time, it dynamically optimizes the threshold according to data quality, improves the sensitivity of small samples and the stringency of large samples, reduces bias, and enhances analysis reliability.

[0023] Other advantages, objectives and features of the present invention will be reflected in part from the following description and will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flowchart of data reception and processing for a technical solution of the present invention; Figure 2 A flowchart of data statistics, calculation and correlation characterization of a technical solution of the present invention; Figure 3 The figure shows the distribution of different sites (CpG and GpC) of normal (green) and tumor (red) samples after dimensionality reduction in an example of the present invention. Figure 4 This is a heat map display of normal and tumor samples at different sites (CpG and GpC) in one example of the present invention; Figure 5 This is an enrichment analysis diagram of different KEGG pathways under specific comparison conditions in an example of the present invention; Figure 6 This is a chromosome site information circle diagram based on the hg19 genome in an example of the present invention; Figure 7 This is a gene-function network diagram based on differential enrichment analysis in an example of the present invention. DETAILED DESCRIPTION

[0025] The present invention will be described in further detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0026] It should be understood that terms such as “having”, “including” and “comprising” used herein do not preclude the existence or addition of one or more other elements or combinations thereof.

[0027] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0028] The present invention provides a DNA methylation and chromatin openness data analysis system, comprising: The data receiving and processing module can use Bismark software to process the sequencing data and generate DNA methylation and chromatin open site data. Specifically, the data receiving and processing module receives the DNA sequencing files of biological samples, which are compatible with data generated by various technologies such as bisulfite sequencing (methylation analysis) and ATAC-seq (open analysis). The sequencing files are filtered by fastp software to remove sequencing adapters and low base quality fragments to ensure the reliability of subsequent analysis data. The filtered sequencing files (fq format) are compared with the reference genome through the bismark module. Due to PCR amplification, there may be sequences that are aligned to the same position in the genome. The icate_bismark module performs deduplication processing to remove repeated sequences generated by PCR amplification to prevent data redundancy from interfering with analysis results. The bismark_methylation_extractor module extracts methylation site information and classifies the obtained site information. The coverage2cytosine module converts the methylation coverage data into a genome-wide cytosine methylation report. The parameter --nome-seq is set to distinguish CpG methylation sites from GpC access sites. A site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation counts, unmethylation counts, access counts, and non-access counts is generated to provide basic data for subsequent analysis. The data statistics and calculation module supports users to specify any physical chromosome segment (such as a gene promoter region, a known enhancer region), or automatically scan a sliding window across the entire genome to screen all methylation and accessibility sites within the target region. Specifically, based on the site report file, it scans a sample for the user-specified physical chromosome segment to generate methylation site data and accessibility data, screens the methylation site data and accessibility site data within the segment, and counts the methylation number + non-methylation number, and the accessibility number + non-accessibility number, and calculates the methylation ratio and chromatin accessibility ratio, respectively, to reflect the methylation level and chromatin accessibility. The data association characterization module fills the NA values ​​in the methylation ratio and openness ratio. It can use the average value of the row or column to fill, or specify a number to fill the NA value first and then calculate the average value of the row or column to fill. The filled methylation ratio and openness ratio data are input into the MOFA model for dimensionality reduction analysis to generate a dimensionality reduction graph that characterizes the clustering relationship between samples and a heat map that shows the correlation between methylation and openness. The correlation between methylation and chromatin openness of the specified chromosome physical segment is judged. The dimensionality reduction graph shows whether methylation or openness can be clustered. If the samples are obviously clustered by methylation or openness, it means that the characteristics of the target area can distinguish samples (such as tumors and normal tissues). The heat map shows the coordinated changes in methylation and openness. If the heat map displayed by the specified chromosome physical segment information is significantly different, it means that methylation and openness are related. If one of the groups is NA value or 0 value, then there is no relationship.

[0029] In the above technical solution, the DNA methylation and chromatin accessibility data analysis system of the present invention integrates DNA methylation and chromatin accessibility data generation through a modular process, realizes the full-process analysis from raw sequencing data to biological significance mining, and reveals the coordinated regulatory relationship between DNA methylation and chromatin accessibility through dimensionality reduction clustering and heat map visualization, reduces manual intervention, greatly improves processing efficiency, and enhances analysis efficiency and biological discovery capabilities.

[0030] like Figure 1-2 As shown, the DNA methylation and chromatin accessibility data analysis method includes the following steps: The DNA sequencing files of biological samples are filtered and quality controlled using fastp software. Low-quality reads and adapter sequences are removed based on preset quality and length thresholds to generate filtered high-quality sequencing files. The filtered sequencing file (fq format) is aligned with the reference genome using the bismark module. By identifying sequence features of methylation sites (such as C→T transitions), the read segments are accurately located to the genome and a BAM file containing the alignment information is generated. The BAM file is deduplicated using the deduplicated_bismark module to remove duplicate sequences generated by PCR amplification. This module identifies duplicate reads by comparing the starting position and sequence consistency, retaining only unique alignments to generate a deduplicated BAM file. The bismark_methylation_extractor module is used to extract methylation site information from the deduplicated BAM file, generating a COV file containing the methylation coverage of each CpG site and recording the number of methylated and unmethylated reads at each site. The COV file was processed and the obtained site information was classified. The methylation coverage data was converted into a genome-wide cytosine methylation report using the coverage2cytosine module. The parameter --nome-seq was set to distinguish CpG methylation sites from GpC access sites. A site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation counts, unmethylation counts, access counts, and non-access counts was generated. The storage format of methylation and access data was unified to facilitate subsequent integrated analysis. Based on the site report file, the methylation number and the openness number are counted. For the specified chromosome physical segment, the methylation site data and the openness site data falling within the segment are filtered from the site report file, and the methylation ratio (methylation number / total number of reads) and chromatin openness ratio (openness number / total number of reads) are calculated respectively; NA values ​​in the methylation ratio and openness ratio are filled. The NA values ​​can be filled with the average value of the row or column, or a number is specified to fill the NA value first and then the average value of the row or column is calculated. The filled methylation ratio and openness ratio data are input into the MOFA model for dimensionality reduction analysis. A dimensionality reduction graph is generated to characterize the clustering relationship of methylation or openness between samples, and a heat map is generated to show the correlation between methylation and openness. The dimensionality reduction graph shows whether methylation or openness can be clustered. If the samples are obviously clustered according to methylation or openness, it means that the characteristics of the target region can distinguish samples (such as tumors and normal tissues). The heat map shows the coordinated changes of methylation and openness. If the heat map displayed by the specified chromosome physical segment information is significantly different, it means that methylation and openness are related. If one of the groups is NA value or 0 value, then there is no relationship. Based on the analysis results of the dimensionality reduction graph and heat map, we identify sample groups with significant clustering and highly correlated chromosomal regions, determine the correlation between methylation of specified chromosomal physical segments and chromatin accessibility, verify the correlation between methylation and accessibility through statistical tests (such as the Pearson correlation coefficient), screen out key areas of coordinated regulation, and provide target candidate regions for subsequent functional verification (such as gene knockout experiments).

[0031] In the above technical solution, the DNA methylation and chromatin openness data analysis method of the present invention connects the methylation and openness data analysis processes in series through a modular design, avoids the data format incompatibility problem caused by traditional multi-tool switching, significantly improves the analysis efficiency, integrates methylation and openness data, and realizes multidimensional analysis of epigenetic regulation, providing an efficient solution for gene expression regulation and disease mechanism research. Multi-step filtering, deduplication and deep screening eliminate low-quality data and technical deviations, controllable data quality, and improve analysis reliability. It supports user-defined target chromosome regions (such as genes or regulatory elements of interest) and is suitable for basic research (such as developmental mechanisms) and applied research (such as cancer marker screening). Combining regional correlation and functional annotation to fill in missing values ​​avoids deviations caused by simple filling and retains key epigenetic signals. The dimensionality reduction graph intuitively displays the sample grouping, and the heat map clearly presents the association pattern between methylation and openness, helping researchers quickly lock in key regulatory regions.

[0032] In a development instance, such as Figure 2 The figure below illustrates the data analysis workflow for processing sequencing files into site information. Data input includes: DNA methylation and chromatin accessibility site information, sample data, chromosome physical segment data, and site matrix data. Functionality includes converting site information to a site matrix, quality control, NA value filling, and sample type-based dimensionality reduction. The goal is to infer the relationship between DNA methylation and chromatin accessibility for selected chromosome physical segments.

[0033] In an example, SRP295352 in the GEO database is used for analysis. This data set can be found in the GEO database, and the five thousand rows of chromosome physical fragments of enhancer are selected here. The site information of the chromosome region is integrated to obtain the site information matrix (the rows are chromosome regions, the columns are sample information, and the sample information is three columns: meth, UNmeth, and methlevel). First, we select the sample data of the methlevel column for MOFA dimensionality reduction (you can choose to fill in NA values ​​before reducing the dimension or not fill in NA values). Several methods of filling in NA values ​​are listed here: calculate the row average value to fill, use numbers instead of NA values ​​to calculate the row average value and then fill. For example Figure 3 As shown in the figure, under the CpG-related data dimension, the characteristics of normal and tumor samples are quite different, and UMAP dimensionality reduction can effectively distinguish the two types of samples. Under the GpC-related data dimension, the characteristics of the two types of samples are less distinguishable, and UMAP dimensionality reduction cannot effectively separate them. That is, the difference in GpC-related features between normal and tumor samples is not significant. Figure 3It is obvious that DNA methylation can be divided into two groups, corresponding to normal and diseased, while chromatin openness cannot be divided into two groups. It can be seen that there is no obvious difference between normal and diseased chromatin openness and it cannot be divided into two groups. For the methlevel column, the NA values ​​were counted and sorted, and the rows with the smallest NA value ratio were selected. The data were displayed in a heat map, as shown in the figure below. Figure 4 As shown, the proportion of DNA methylation and chromatin openness is shown. The left figure shows the CpG site methylation pattern of normal (light green) and tumor (orange) samples. The color from red to blue indicates the methylation level from high to low. The right figure shows the GpC site openness pattern of normal (light green) and tumor (orange) samples. The color from red to blue indicates the openness level from high to low (here we mark the NA value in white for easy distinction). Then select the same part of the chromosome physical segment to view the relationship between DNA methylation and chromatin openness. At the same time, for the chromosome physical segment, differential analysis was performed to obtain p-value and logFC value, and GSEA gene set enrichment analysis data was obtained to draw a differential analysis curve, as shown below. Figure 5 As shown, the horizontal axis is the ranking of genes by contrast value, and the vertical axis is the gene enrichment score, which reflects the enrichment degree of the pathway in the ranking. Different colored lines represent different pathways. Statistical methods such as rank sum test and t.test are used to calculate p-value and logFC to evaluate the significance of enrichment. Here, we check the enriched pathways with p-value less than 0.1 in the chromosome region. For the site information obtained at the beginning, in order to facilitate the viewing of methylation status, such as Figure 6 As shown in the figure, a circular diagram is drawn based on hg19 to display the site information on the chromosome (such as methylation status) through different colors and regions. The circular structure makes it easy to intuitively present the distribution characteristics of the site on the chromosome, helping to quickly locate and compare the properties of different regions. Differential analysis and enrichment analysis were performed on the physical chromosome fragments input at the beginning. In order to facilitate the viewing of the relationship between related genes and pathways, as shown in the figure, Figure 7 As shown, a gene-pathway network diagram was drawn based on the KEGG database.

[0034] In another technical solution, the NA values ​​in the methylation ratio and the openness ratio are filled, and the filling method is specifically as follows: A dynamic threshold judgment module was set up, with a pre-set threshold for the NA value ratio (10%) and a sequencing depth threshold (30X). The module read the methylation site report file, scanned the NA value ratio of each chromosome segment line by line, and selected a filling strategy based on the degree of data missing, that is, the NA value ratio of each line in the methylation site report file; When the NA value ratio of a row is ≤10%, for low-deletion rows, the system detects the sequencing depth (calculated by the BAM alignment file) and genomic functional annotation information (obtained from databases such as UCSC and ENCODE) of the chromosome fragment corresponding to the row, retaining local biological signals in high-confidence regions and avoiding over-reliance on single-site data in low-quality regions. Specifically: If the sequencing depth is ≥30X and the chromosome fragment is located in the CpG island or gene promoter annotation region, that is, the high-depth functional region is filled with the weighted sliding mean of the sequencing depth of adjacent sites. The weight is calculated by the ratio of the sequencing depth of the current site to the sum of the sequencing depths of all sites in the sliding window. For example, with the current site as the center, sites within 50bp upstream and downstream are assigned weights based on the sequencing depth (for example, if the depth of site A is 30X and the depth of site B is 40X, the weights are 30 / 70 and 40 / 70, respectively), and the weighted mean is calculated; If the sequencing depth is <30X or located in a non-functional annotation region, that is, the low-depth or non-functional region is directly filled with the sliding window mean of the adjacent sites.

[0035] When the NA value ratio of a row is greater than 10%, for highly missing rows, the system queries whether the chromosome segment belongs to a pre-annotated enhancer or conserved region (through PhastCons score or ENCODE annotation). The biological conservation of functional regions is used to improve the filling reliability and avoid the global median from masking local features. Specifically: If it belongs to an enhancer or conserved region, that is, a functional annotation matching region, the median methylation number of all samples in the same functional category (such as enhancer) is used for filling. For example, if the region belongs to an enhancer, the median methylation number of the enhancer region in all samples is taken; If the functional annotation cannot be matched, the median of the global methylation number of the physical segment of the chromosome is used for filling (such as the median value of the entire chromosome 1p36 region); After the filling is completed, the system generates a filling report, marking the filling strategy and data source of each region, and inputs the filled methylation data into the subsequent correlation analysis module for MOFA dimensionality reduction.

[0036] In the above technical solution, the accuracy and biological rationality of NA value filling are achieved through data quality judgment-functional annotation matching-strategy intelligent selection. Different filling strategies are matched for different NA missing degrees and data scenarios to avoid the deviation caused by "one size fits all". The dynamic layered filling strategy takes into account both local signals and global trends to improve data integrity. For high-functional regions such as CpG islands and promoters, sequencing depth weighted filling is used to avoid noise interference in low-depth areas and reduce functional region deviations. Through functional annotation-driven median filling, the apparent characteristics of conserved regions are preferentially retained to enhance biological rationality. The dynamic threshold design can flexibly respond to samples with different sequencing depths and data missing degrees, and adapt to complex data scenarios.

[0037] In another technical solution, the filled methylation ratio and openness ratio data are subjected to dimensionality reduction analysis to generate a dimensionality reduction graph to characterize the clustering relationship of methylation or openness between samples, and a heat map is generated to show the correlation between methylation and openness, specifically: A dynamic weight allocation module was set up in the MOFA model to quantify the discreteness of the two types of data by calculating the coefficient of variation (CV value) of the methylation ratio and the openness ratio of each sample. The weights of the two types of data in dimensionality reduction were adjusted according to the coefficient of variation of the methylation ratio and the openness ratio. Specifically, The coefficient of variation of the methylation ratio and the coefficient of variation of the openness ratio were calculated for each sample, i.e., the standard deviation of the methylation (or openness) ratio divided by its mean, reflecting the relative fluctuation intensity of the data between samples; If the coefficient of variation of the methylation ratio is higher than that of the openness ratio, the weight of the methylation data is 1.5 times that of the openness data; if the coefficient of variation of the openness ratio is higher than that of the methylation ratio, the weight of the openness data is 1.5 times that of the methylation data; the weight is dynamically adjusted according to the intrinsic heterogeneity of the data to ensure that high-variance data dominates the dimensionality reduction and enhance the ability to analyze sample differences.

[0038] The data after dynamic weight assignment is input into the MOFA model to generate a dimensionality reduction graph. The weight distribution areas are marked in the graph to indicate the main sources of cluster differences, intuitively displaying the sample clustering relationship and clearly indicating which data (methylation or openness) dominates the cluster differences. Based on the dimensionality reduction results, the Pearson correlation coefficient matrix of the methylation and openness ratios of all chromosomal regions was calculated, and a correlation coefficient heat map of the methylation and openness ratios was generated. The weight distribution results were combined with the heat map correlation analysis to screen out chromosomal physical segments with significant weight differences (such as weight ratio >1.3 times) as key areas for the coordinated regulation of methylation and openness. By combining statistics and visualization, the potential functional areas for the dynamic coupling of methylation and openness were identified.

[0039] In the above technical solution, the dynamic weight allocation mechanism avoids setting fixed weights for methylation and openness data, but automatically adjusts the weights according to the degree of variation of the data itself, more flexibly reflecting the relative importance of the two types of data in different samples, strengthening the contribution of high-variation data, and improving clustering resolution. By emphasizing the role of data with high variation in dimensionality reduction, samples can be more effectively clustered according to their methylation and openness characteristics. In the dimensionality reduction graph, the clustering structure of the samples is clearer, which can better distinguish different sample groups and help discover potential biological subgroups or differences between samples. Marking the weight distribution area in the dimensionality reduction graph provides researchers with an intuitive visualization tool to intuitively display the contribution of methylation or openness to clustering. It can be applied to various biological research scenarios and has wide applicability and versatility.

[0040] In another technical solution, traditional weight assignment ignores the differences in relationships between samples, resulting in an inadequate representation of the data's intrinsic structure. The system calculates similarity between samples for both methylation ratios and accessibility ratios. In the dynamic weight assignment module, the weight assignment is dynamically adjusted based on the similarity matrix of methylation ratios and accessibility ratios between samples. By calculating the Euclidean distance or correlation coefficient between samples, the differences in epigenetic patterns between samples are fully captured. The Euclidean distance measures the absolute difference in methylation or accessibility ratios between samples, and the correlation coefficient measures the consistency of the change trend between samples. Independent similarity matrices for methylation data and accessibility data are generated. The similarity matrices of methylation and accessibility data are hierarchically clustered to generate cluster dendrograms. The silhouette coefficient is used to quantify the clustering tightness. The closer the value is to 1, the tighter the samples of the same type are and the more separated the samples of different types are. If the clustering tightness of the methylation data similarity matrix is ​​higher than that of the openness data, the weight of the methylation data is increased; if the clustering tightness of the openness data similarity matrix is ​​higher than that of the methylation data, the weight of the openness data is increased. The weights are dynamically optimized based on the intrinsic structure of the data, strengthening the data types that contribute more to sample clustering. The adjusted weights are input into the MOFA model for dimensionality reduction, and the regions where high-weight data contribute to the clustering results are marked in the graph using color gradients or symbol size, visually demonstrating which data types dominate the sample clustering and locating their corresponding chromosomal regions. Combining dimensionality reduction graphs with heat map analysis, chromosomal segments with significant correlations between methylation and openness in high-weight regions are screened and marked as candidate regulatory regions. Their biological significance is verified using functional annotation databases (such as GO and KEGG).

[0041] In the above technical solution, through similarity analysis-closeness comparison-dynamic weight optimization, the intelligent integration of methylation and openness data is achieved, making the dimensionality reduction results closer to the real biological signals. The weight allocation is based on the similarity and clustering effect of the data itself, rather than subjective pre-setting, and can adapt to the characteristic differences of different sample cohorts. By strengthening the data features with high clustering closeness, the sample grouping in the dimensionality reduction diagram is clearer and more distinguishable. The marking of the weight contribution area provides researchers with intuitive clues to quickly locate the key data type (methylation or openness) that drives sample clustering. Avoiding the "averaging" defect of traditional fixed weights, dynamic adjustment achieves the complementary advantages of methylation and openness data.

[0042] In another technical solution, a single weight allocation cannot adapt to the local data characteristics, resulting in insufficient stability of the dimensionality reduction results. This problem is solved through phased iterative optimization.

[0043] The dimensionality reduction process is divided into two stages: preliminary dimensionality reduction and iterative optimization; In the initial dimensionality reduction stage, the MOFA model is run with default weights to generate an initial dimensionality reduction graph. A clustering algorithm (such as DBSCAN or K-means) is used to identify preliminary cluster groups and mark the main clusters. In the iterative optimization phase, the weights are redistributed based on the degree of dispersion of the methylation and openness data within the cluster groups, evaluated by the distribution range or density of the data points within the group. For example, the initial weight is set to 1:1, and the adjustment range is dynamically set each time based on the difference in dispersion, and the dimensionality reduction results are updated: If the degree of dispersion of methylation data in a certain group is higher than that of open data, the weight of methylation data in this group will be reduced; if the degree of dispersion of open data is higher than that of methylation data, the weight of open data in this group will be reduced. By adjusting the local weights, the interference of highly discrete data on the overall dimensionality reduction results can be suppressed, thus improving clustering stability. The adjusted weights are re-input into the MOFA model to generate an updated dimensionality reduction graph, and the above process is repeated (usually 2-3 times) until the clustering results tend to be stable (such as the clustering similarity of two consecutive iterations is greater than 90%). Through multiple optimizations, the optimal weight distribution is gradually approached to avoid the limitations of a single adjustment.

[0044] Output the final dimensionality reduction graph and use color gradients or arrows to mark the clustering change trajectory before and after weight adjustment to intuitively show the impact of weight adjustment on sample clustering.

[0045] In the above technical solution, through phased dimensionality reduction and iterative weight optimization, the weights of methylation and openness data can be dynamically adjusted according to the characteristics of different cluster groups, thereby achieving refined analysis of the data. The weights are automatically adjusted according to the degree of discreteness of the data within each cluster group, so it has strong adaptability. In the iterative optimization process, by continuously adjusting the weights, some potential information in the data that was originally hidden can be mined. For example, in some cases, the initial dimensionality reduction graph may not clearly show the differences between certain samples, but through weight adjustment and iterative optimization, these differences may gradually emerge, which helps to discover new biological phenomena or subgroup structures and provide clues for further research. Iterative optimization usually only takes 2-3 rounds to converge. The process of multiple rounds of iterative optimization can make the dimensionality reduction results more stable and can better reveal the intrinsic structure and biological characteristics of methylation and openness data.

[0046] In another technical solution, traditional dimensionality reduction methods cannot capture dynamic changes in time, resulting in the lack of time series correlation analysis. By capturing time series and adjusting dynamic weights, the interference of abnormal time points can be effectively suppressed.

[0047] Integrating time series data into the MOFA model, the methylation ratio and accessibility ratio at different time points (e.g., developmental stage, disease progression) are input as independent views. The data at each time point are independently preprocessed and standardized to ensure consistency across the time dimension. At each time point, the trajectory of epigenetic changes over time is captured to avoid data confusion across time points. The correlation between the methylation change trend (such as the difference between adjacent time points) and the openness change trend within that time point is calculated. The weight of the data at each time point is dynamically adjusted based on the consistency of the change trend between time points. If the methylation and openness change trends at a certain time point are highly consistent (e.g., correlation coefficient ≥ 0.8), it is marked as a "cooperative time point" and the data at that time point is given a higher weight. If the trends are inconsistent (e.g., correlation coefficient ≤ 0.2), the weight is reduced or the time point is marked as an abnormal time point (e.g., an outlier caused by experimental error), thereby strengthening the data contribution of key time points and weakening the influence of abnormal or noisy time points. Based on the dynamic weights, the weighted MOFA model is run to output the dimensionality reduction coordinates of each time point. By connecting the coordinates of adjacent time points, a dimensionality reduction trajectory diagram across time points is generated, showing the co-evolutionary relationship between methylation and openness over time, and intuitively presenting the dynamic evolution of the apparent state under disease progression or therapeutic intervention.

[0048] In the above technical solution, through time series modeling → trend consistency weighting → trajectory visualization, the temporal laws of coordinated regulation are revealed, the turning points of methylation-openness coordinated changes are automatically identified, and the noise time points are filtered through the weight mechanism. Based on the trend judgment of biological logic, experimental errors or non-specific changes are eliminated, and the trajectory diagram intuitively displays the epigenetic history of the sample.

[0049] In another technical solution, a visualization interface is developed in the R Shiny or Python Dash framework, integrating the dimensionality reduction graph, heat map, and weight control panel. Users can interact with the graph by dragging, clicking, and other operations. The interactive visualization interface is embedded, allowing users to manually adjust the weight parameters based on the preliminary results of the dimensionality reduction graph. Provide independent sliders for methylation and openness weights. The slider controls dynamically modify the weight ratio of methylation and openness data. After the user adjusts, the system calls the MOFA model to recalculate and update the dimensionality reduction graph in real time. When the user selects a specific clustering area, the system automatically calculates the weight recommendations for the two types of data in the area and prompts whether to adopt them. Through human-computer collaborative optimization, the optimal weight configuration is quickly located. The user's modified weight parameters are fed back to the MOFA model to regenerate the dimensionality reduction graph, and the adjustment history is recorded for retrospection. The system automatically records the parameters and dimensionality reduction results of each weight adjustment, supports users to trace back to any historical state, and export charts and data of key nodes to facilitate comparison of the effects of different weight strategies and optimize the analysis process.

[0050] In this technical solution, users can observe the impact of weight changes on clustering results through intuitive sliders, eliminating the need to rely on complex algorithmic parameter adjustments. Experimenters can adjust weights based on biological experience (e.g., manually increasing the methylation weight if abnormal methylation of a gene promoter is known to be associated with a disease) to validate hypotheses and guide analysis. A historical backtracking function ensures reproducible analysis, avoiding inconsistent results due to forgotten parameters and complying with scientific research data management standards.

[0051] In another technical solution, an automated enrichment analysis module is set up to integrate the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. For significantly different chromosome physical segments (such as segments with p-values ​​< 0.05), functional matching is performed using the gene function annotation database, such as coordinate matching and semantic analysis, to automatically associate biological function entries. The genomic coordinates of the chromosome fragments were compared with the gene coordinates in the database. If the fragments covered the promoter, exon, or regulatory region of the gene, the functional annotation of the gene was associated. The hypergeometric test or Fisher's exact test was used to calculate the enrichment significance (FDR-corrected p-value) of the successfully matched functional entries. The entries with FDR < 0.1 were screened to identify pathways or biological processes related to the coordinated regulation of methylation and accessibility. Generate an association enrichment map, using chromosome physical segments as nodes and functional annotations as labels, annotate the synchronous change direction of methylation ratio and openness ratio within the segment, and annotate the synchronous change direction of methylation and openness ratio of the segment. If multiple segments are associated with the same functional entry, they are connected by edges to form a functional cluster. Based on the enrichment map, a list of high-confidence co-regulatory regions is output. The regions must meet the following conditions at the same time: 1) Functional significance: matching at least one significantly enriched GO term or KEGG pathway (FDR < 0.1); 2) Apparent synergy: The absolute value of the Pearson correlation coefficient of the methylation and accessibility ratios is ≥ 0.6, ensuring that the changes in the two are highly correlated; 3) Functional element verification: The region covers at least one known enhancer or transcription factor binding site; Through multi-condition joint screening, regions with random associations or low functional value are excluded to enhance the biological credibility of the results. The list of co-regulated regions is superimposed on the original dimensionality reduction map, and their distribution characteristics in the sample cluster are displayed through color or shape markings. The consistency of their spatial distribution is verified by chromosome overlap heat map, and the density heat map of methylation and openness ratio is drawn along the chromosome coordinate axis to verify the spatial clustering of high-confidence regions on the genome.

[0052] In this technical solution, the system automatically maps chromosome segments to functional entries, eliminating the need for manual database searches. This system directly links epigenetic differences to gene function, significantly improving analysis efficiency. Three criteria ensure that the selected regions are not only statistically significant but also possess real biological functions: functional enrichment avoids interference from meaningless regions, high correlation ensures that the coordinated changes in methylation and accessibility are not accidental, and regulatory element coverage targets regions with potential regulatory effects (e.g., hypomethylation and high accessibility in enhancer regions often activate gene expression). The enrichment map intuitively displays the association between chromosome segments and functional pathways, forming a visual functional network.

[0053] In another technical solution, the problems of missed detection in small samples or low-depth data and high false positive rate in high-quality data caused by traditional fixed thresholds are dynamically adjusted through adaptive enrichment significance thresholds.

[0054] The system automatically counts the number of samples in the input data (distinguishing between small samples ≤ 30, medium samples 31-99, and large samples ≥ 100), calculates the average sequencing depth of each site using the alignment file (BAM) (distinguishing between low depth < 20X, medium depth 20X-49X, and high depth ≥ 50X), and obtains the sample size and sequencing depth. Based on the sample size and sequencing depth of the input data, the adaptive enrichment significance threshold is automatically calculated. Small samples or low-depth data have insufficient statistical power, and strict thresholds (such as FDR < 0.05) can easily lead to missed detection of key signals. Therefore, if the sample size is ≤ 30 or the average sequencing depth is < 20X, the FDR threshold can be relaxed to < 0.2; If the sample size is ≥100 and the average sequencing depth is ≥50X, the FDR threshold is tightened to <0.01, and the screening criteria are improved to reduce false positives; If the sample size is 31-99 or the average sequencing depth is 20X-49X, the standard threshold (such as FDR < 0.05) is used by default.

[0055] In the above technical solution, the adaptive enrichment significance threshold is dynamically adjusted according to the data quality, which improves the analytical sensitivity of small sample data, enhances the rigor of high-depth data, and improves the robustness of adaptive data quality, thereby reducing subjective bias.

[0056] The number of devices and processing scales described herein are intended to simplify the description of the present invention. Applications, modifications, and variations of the present invention will be readily apparent to those skilled in the art.

[0057] Although the embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the description and implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. DNA methylation and chromatin openness data analysis system, characterized by: include: The data receiving and processing module receives the DNA sequencing file of the biological sample, performs filtering, compares the filtered sequencing file with the reference genome, removes duplicates, extracts methylation site information, sets the parameter --nome-seq, distinguishes CpG methylation sites from GpC open sites, and generates a site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation counts, unmethylation counts, open counts, and non-open counts; The data statistics and calculation module, based on the site report file, selects the methylation site data and open site data falling within the user-specified chromosome physical segment, and calculates the methylation ratio and chromatin openness ratio respectively; The data association characterization module fills the NA values ​​in the methylation ratio and the openness ratio, inputs the filled methylation ratio and openness ratio data into the MOFA model for dimensionality reduction analysis, generates a dimensionality reduction graph representing the clustering relationship between samples and a heat map showing the correlation between methylation and openness, and determines the correlation between methylation and chromatin openness of the specified chromosome physical segment.

2. A method for analyzing DNA methylation and chromatin accessibility data, characterized in that: The following steps are involved: Filtering the DNA sequencing file of the biological sample to generate a filtered sequencing file; Compare the filtered sequencing file with the reference genome to generate a BAM file; Deduplication is performed on the BAM file to generate a deduplicated BAM file; Extract methylation site information from the deduplicated BAM file and generate a COV file of methylation coverage; Process the COV file and set the parameter --nome-seq to distinguish CpG methylation sites from GpC open sites, and generate a site report file containing chromosomes, chromosome start sites, chromosome end sites, methylation numbers, unmethylation numbers, open numbers, and non-open numbers; Based on the site report file, the methylation number and the openness number are counted. For the specified chromosome physical segment, the methylation site data and the open site data falling within the segment are screened to calculate the methylation ratio and chromatin openness ratio respectively; Fill the NA values ​​in the methylation ratio and openness ratio, perform dimensionality reduction analysis on the filled methylation ratio and openness ratio data, generate a dimensionality reduction graph to characterize the clustering relationship of methylation or openness between samples, and generate a heat map to show the correlation between methylation and openness; Based on the analysis results of the dimensionality reduction graph and heat map, the correlation between the methylation of the specified chromosome physical segment and the chromatin openness was determined.

3. The DNA methylation and chromatin accessibility data analysis method according to claim 2, wherein: Fill the NA values ​​in the methylation ratio and the openness ratio. The filling method is as follows: Set up a dynamic threshold judgment module and select a filling strategy based on the proportion of NA values ​​in each row of the methylation site report file; When the NA value ratio of a row is ≤10%, the sequencing depth and genome functional annotation information of the chromosome fragment corresponding to the row are detected; If the sequencing depth is ≥30X and the chromosome fragment is located in the CpG island or gene promoter annotated region, the weighted sliding mean of the sequencing depth of adjacent sites is used for filling, where the weight is calculated by the ratio of the sequencing depth of the current site to the sum of the sequencing depths of all sites in the sliding window; If the sequencing depth is <30X or located in a non-functional annotation region, the sliding window mean of the adjacent sites is directly used for filling; When the NA value of a row accounts for >10%, query whether the chromosome segment belongs to the pre-annotated enhancer or conserved region; If it belongs to an enhancer or conserved region, the median methylation number of all samples in the same functional category region is used for filling; If the functional annotation cannot be matched, the median global methylation number of the physical segment of the chromosome is used for filling; The filled methylation data were input into the subsequent association analysis module.

4. The method for analyzing DNA methylation and chromatin accessibility data according to claim 2, wherein: The filled methylation ratio and openness ratio data were subjected to dimensionality reduction analysis to generate a dimensionality reduction graph to characterize the clustering relationship of methylation or openness between samples, and a heat map was generated to show the correlation between methylation and openness, specifically: A dynamic weight allocation module is set up in the MOFA model to adjust the weights of the two types of data in dimensionality reduction according to the coefficient of variation of the methylation ratio and the openness ratio; The coefficient of variation of the methylation ratio and the coefficient of variation of the openness ratio were calculated for each sample; If the coefficient of variation of the methylation ratio is higher than the coefficient of variation of the openness ratio, the weight assigned to the methylation data is 1.5 times that of the openness data; If the coefficient of variation of the openness ratio is higher than the coefficient of variation of the methylation ratio, the weight assigned to the openness data is 1.5 times that of the methylation data; Generate a dimensionality reduction graph based on the data after dynamic weight assignment, and mark the weight distribution area in the graph to indicate the main source of cluster differences; The weight distribution results were combined with heat map correlation analysis to screen out chromosome physical segments with significant weight differences as key areas for the coordinated regulation of methylation and openness.

5. The method for analyzing DNA methylation and chromatin accessibility data according to claim 4, wherein: In the dynamic weight assignment module, the weight assignment is dynamically adjusted based on the similarity matrix of methylation ratios and openness ratios between samples; By calculating the Euclidean distance or correlation coefficient between samples, independent similarity matrices for methylation data and openness data were generated; If the clustering tightness of the similarity matrix of methylation data is higher than that of open data, the weight of methylation data is increased; If the clustering tightness of the similarity matrix of the open data is higher than that of the methylation data, the weight of the open data is increased; The adjusted weights are input into the MOFA model for dimensionality reduction, and the areas where high-weight data contribute to the clustering results are marked.

6. The method for analyzing DNA methylation and chromatin accessibility data according to claim 4, wherein: The dimensionality reduction process is divided into two stages: preliminary dimensionality reduction and iterative optimization; In the initial dimensionality reduction stage, the default weights are used to generate the initial dimensionality reduction graph and identify the preliminary clustering groups; In the iterative optimization stage, the weights are redistributed and the dimensionality reduction results are updated according to the discrete degree of methylation and openness data within the cluster group; If the dispersion of methylation data in a group is higher than that of open data, the weight of methylation data in the group will be reduced; If the dispersion of open data is higher than that of methylation data, the weight of open data in this group is reduced; Output the final dimensionality reduction graph and mark the clustering change trajectory before and after weight adjustment.

7. The method for analyzing DNA methylation and chromatin accessibility data according to claim 4, wherein: Integrate time series data into the MOFA model, inputting methylation ratios and openness ratios at different time points as independent views; Dynamically adjust the weight of data at each time point based on the consistency of change trends between time points; If the methylation at a certain time point is highly consistent with the trend of openness change, the data at that time point will be given a higher weight; If the trend is inconsistent, reduce the weight or mark it as an abnormal time point; Generate dimensionality-reduced trajectory plots across time points to demonstrate the co-evolutionary relationship between methylation and openness over time.

8. The method for analyzing DNA methylation and chromatin accessibility data according to claim 4, wherein: Embed an interactive visualization interface that allows users to manually adjust weight parameters based on the preliminary results of the dimensionality reduction graph; Provide a slider control to dynamically modify the weight ratio of methylation and openness data, and update the dimensionality reduction graph in real time; When the user selects a specific clustering area, the weighted values ​​of the two types of data in the area are automatically calculated and the user is prompted to decide whether to adopt them. The user-corrected weight parameters are fed back to the MOFA model to regenerate the dimensionality reduction graph, and the adjustment history is recorded for retrospective use.

9. The method for analyzing DNA methylation and chromatin accessibility data according to claim 2, wherein: An automated enrichment analysis module was set up to perform functional matching on significantly different chromosome physical segments using gene function annotation databases, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG); Calculate the enrichment significance of the successfully matched functional entries to screen out pathways or biological processes related to the coordinated regulation of methylation and openness; Generate a correlation enrichment map, using chromosome physical segments as nodes and functional annotations as labels, and mark the synchronous change direction of methylation ratio and openness ratio within the segment; Based on the enrichment map, a list of high-confidence co-regulatory regions is output. The regions must meet the following conditions at the same time: 1) Match at least one significantly enriched GO term or KEGG pathway; 2) the absolute value of the Pearson correlation coefficient of the methylation-openness ratio was ≥0.6; 3) The region covers at least one known enhancer or transcription factor binding site; The list of co-regulatory regions was superimposed on the original dimensionality reduction map, and the consistency of their spatial distribution was verified by chromosome overlap heat map.

10. The method for analyzing DNA methylation and chromatin accessibility data according to claim 9, wherein: Automatically calculate the adaptive enrichment significance threshold based on the sample size and sequencing depth of the input data; If the sample size is ≤30 or the average sequencing depth is <20X, the FDR threshold is relaxed to <0.2; If the sample size is ≥100 and the average sequencing depth is ≥50X, the FDR threshold is tightened to <0.01.

Images