Hi-C data quality rapid evaluation method Motif-Hi-C based on adjacently connected Motif sequence and application of Hi-C data quality rapid evaluation method Motif-Hi-C

Abstract

The invention discloses a Hi-C data quality evaluation method Motif-Hi-C based on an adjacent connection Motif sequence. The Hi-C data quality evaluation method comprises the following steps: S1, carrying out Motif matching on a fastq file stored by Hi-C data; s2, classifying the original Hi-C data storage fastq file: classifying the original Hi-C data storage fastq file into a matched.fastq file and an unmatched.fastq file according to the restriction enzyme cutting site and Motif characteristics of the restriction enzyme, and classifying the original Hi-C data storage fastq file into an original Hi-C data storage fastq file and an original Hi-C data storage fastq file; s3, noise reduction processing of the unmatched.fastq file data is carried out, and after the unmatched.fastq file data are compared, filtered and subjected to duplicate removal, the unmatched.fastq file data subjected to noise reduction and an evaluation result are obtained; s4, carrying out matched.fastq file data simulation and de-duplication on the matched.fastq file data; and S5, Hi-C data quality evaluation: combining a data evaluation result of the unmatched.fastq file subjected to noise reduction processing in the step S3 with a data evaluation result of the matched.fastq file subjected to simulated duplicate removal, calculating a ratio of an effective interaction number to a total reading pair, and evaluating the Hi-C data quality through the ratio. The method provides a new thought and scheme for high-throughput Hi-C sequencing data analysis, and the Hi-C data analysis efficiency can be greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of genome data analysis, and in particular to a method for rapid Hi-C data quality assessment based on proximity-linked Motif sequences, Motif-Hi-C, and its application. Background Art

[0002] High-throughput chromatin conformation capture (Hi-C) technology is based on the 3C technology "proximity ligation". It combines high-throughput sequencing technology with bioinformatics analysis to reveal the three-dimensional structure of the genome on a genome-wide scale. Hi-C technology uses restriction endonucleases to recognize specific restriction sites to cut the genome, and then uses biotin filling and proximity ligation to combine two spatially adjacent but discontinuous DNA sequences into a chimeric fragment (the fragment is the Motif). After constructing a complete Hi-C library and sequencing it, the original Hi-C data is generated, stored in fastq format, and used for subsequent analysis and processing.

[0003] Hi-C technology has strict requirements on experimental operation. Therefore, a large amount of data noise is often generated in Hi-C data sets due to problems with the Hi-C experimental principle and reagent reaction efficiency, such as random fragments, self-linking (Self Cycle, DNA region fragments connect to themselves) and dangling ends (Dangling Ends, the DNA sequence filled in at the end of biotin is not successfully connected to the adjacent end) (Eagen et al., 2018, Trends in Biochemical Sciences). These noises will seriously affect the accuracy of the analysis results (Wang et al., 2018, Nature Plants), so the original fastq files need to be processed. At present, the Hi-C data processing process is mainly divided into two key stages: the upstream stage is to extract Hi-C effective interaction information data from fastq files (Touray et al., 2023, Bio Protocol) (upstream analysis); the downstream stage uses Hi-C effective interaction information data to analyze different chromatin structures, such as the identification and mapping analysis of chromatin three-dimensional structure (spatial conformation), including chromatin territory, chromatin compartment, topologically associated domain (TAD) and chromatin loop (Loop) (downstream analysis). The main goal of the upstream stage of Hi-C data is to remove noise from the raw data and extract effective interaction information. Currently commonly used Hi-C data quality assessment software, such as Hi-C Pro, HiCUP, Juicer, Hicexplorer, etc., can effectively evaluate Hi-C data. Among them, Hi-C Pro provides a complete upstream analysis process (Servant et al., 2015, Genome Biology), HiCUP can more quickly extract effective interaction data from fastq files (Wingett et al., 2015, F1000Research), Juicer can support both upstream and downstream analysis steps (Durand et al., 2016, Cell System), and HiCExplorer has high data quality control capabilities (Wolff et al., 2018, Nucleic Acids Research). Although the commonly used Hi-C data quality assessment software can effectively process Hi-C data, in the case of large Hi-C data sets or low server performance, these software have the problem of taking a long time to process the upstream analysis stage of Hi-C data and occupying a large amount of computing resources, which makes it impossible for researchers to quickly judge the quality of Hi-C data sets.

[0004] Hi-C technology can capture all spatial interaction information within and between chromatin in the whole genome, thus generating a very large and complex data set. For large eukaryotic genomes, the number of theoretical interactions increases exponentially with the size of the genome. At the same time, since the original fastq files obtained by Hi-C experiments contain a lot of data noise, how to efficiently and accurately process these huge Hi-C data to obtain effective interaction information within a reasonable time, reasonable resources and storage requirements is crucial for subsequent analysis.

[0005] Therefore, it is necessary to develop a rapid Hi-C data quality assessment method applicable to large-scale Hi-C data. Summary of the invention

[0006] The purpose of the present invention is to provide a method for rapid assessment of Hi-C data quality based on proximity-linked Motif sequences, Motif-Hi-C, and its application, to solve the above problems.

[0007] According to a first aspect of the present invention, a method for evaluating Hi-C data quality based on proximity-linked Motif sequences, Motif-Hi-C, is provided. The method comprises the following steps: S1: Hi-C data storage fastq file for Motif matching: For single enzyme digestion Hi-C data, the KMP algorithm is used to match the Motif, and for multiple enzyme digestion Hi-C data, the AC automaton algorithm is used to match the Motif; S2: Classify the original Hi-C data storage fastq files: Classify the original Hi-C data storage fastq files into matched.fastq and unmatched.fastq files according to the restriction endonuclease cutting sites and Motif characteristics; S3: Noise reduction processing of unmatched.fastq file data: After performing alignment, filtering and deduplication steps on the unmatched.fastq file data, the noise-reduced unmatched.fastq file data and evaluation results are obtained; S4: simulated deduplication of matched.fastq file data: calculate the PCR amplification deduplication multiple according to the deduplication of unmatched.fastq file data, and then use the PCR amplification deduplication multiple calculated from the unmatched.fastq file data to simulate the PCR amplification deduplication of the matched.fastq file evaluation results; S5: Hi-C data quality assessment: The data assessment results of the unmatched.fastq file after noise reduction in step S3 are combined with the data assessment results of the matched.fastq file after simulated deduplication in step S4, and the ratio of the valid interaction number (valid interaction rmdup) to the total read pairs (total read pairs) is calculated to evaluate the Hi-C data quality.

[0008] Therefore, after classifying Hi-C data, this method uses different strategies for subsequent processing of different types of data, which can greatly improve the efficiency of Hi-C data quality assessment, and this method is efficient, scientific and accurate. The application of Motif-Hi-C provides new ideas and solutions for high-throughput sequencing data analysis, especially for large-scale Hi-C data quality analysis, which can greatly improve analysis efficiency and speed.

[0009] In certain embodiments, in step S2 of the method, the method for classifying the original Hi-C data storage fastq file comprises the following steps: a1: Screening read IDs containing specific motifs: Read the fastq files containing Hi-C sequencing reads. For each read in the R1.fastq and R2.fastq files, the program will search for read IDs containing specific motifs in their IDs; a2: Output qualified read IDs: For reads whose IDs contain specific motifs, the program will output their IDs into two new files respectively; a3: Merge and remove duplicate IDs: The program will merge the two folders in step a2 into a new file and rename it to ensure that all qualified read IDs are integrated. The merged ID list will then be deduplicated and saved as a new file. a4: Classify fastq files according to the ID list: Use the ID list of the file after deduplication in step a3 to classify the original R1.fastq and R2.fastq files. If the read ID exists in the file after deduplication in step a3, the read is classified as matched, and R1.matched.fastq and R2.matched.fastq files are generated for R1.fastq and R2.fastq, respectively; if the read ID does not exist in the file after deduplication in step a3, the read is classified as unmatched, and R1.unmatched.fastq and R2.unmatched.fastq files are generated for R1.fastq and R2.fastq, respectively.

[0010] By classifying the original Hi-C data into two types of data files, matched.fastq and unmatched.fastq, the quality assessment of Hi-C data can be completed. This can greatly reduce the amount of upstream data analysis and improve the speed and efficiency of data analysis. This method has been verified to be accurate and reliable, and is particularly suitable for large-scale Hi-C data quality assessment.

[0011] In certain embodiments, the method of using the KMP algorithm to match Motifs for single enzyme digestion Hi-C data in step S1 of the method comprises the following steps: 1) Build a partial match table based on the Motif sequence to quickly move the pointer during the matching process, and initialize a partial match table with the same length as the Motif, with all initial values ​​being 0; 2) Starting from the second character of the motif, calculate the longest common prefix and suffix length of each position in turn and record them in the partial match table. During the matching process, use the constructed partial match table to search for the motif in the read sequence; 3) Initialize two pointers to point to the start position of the reads sequence and Motif respectively; 4) Move the pointers in the reads sequence in sequence, compare the base characters at the corresponding positions, and adjust the position of the Motif according to the partial match table: if the current base character matches successfully, both pointers move back one position at the same time; if the current base character does not match and the value of the partial match table is greater than 0, move the Motif to the right and keep the pointer of the reads sequence unchanged; if the current base character does not match and the value of the partial match table is 0, move both the Motif and reads sequence pointers to the right one position; 5) Repeat the above process until the match is completed or the pointer of the reads sequence moves to the end.

[0012] The KMP algorithm is a classic string matching algorithm that is suitable for matching a single pattern. For single-enzyme Hi-C data, since only one specific enzyme cutting site (i.e., a specific Motif) is involved, the KMP algorithm can efficiently identify all matching sites without having to process complex pattern combinations. For the processing of single-enzyme Hi-C data, the KMP algorithm helps reduce code complexity and potential errors.

[0013] In certain embodiments, the method for matching Motifs using the AC automaton algorithm for multi-enzyme Hi-C data in step S1 comprises the following steps: 1) Construct a Trie tree as the search data structure of the AC automaton; 2) Construct a fail pointer so that when the current base character fails to match, it jumps to the base character with the longest common prefix and suffix to continue matching; 3) Scan the reads sequence for matching.

[0014] The AC automaton algorithm is an efficient multi-mode string matching algorithm. The algorithm can process multiple modes at the same time and output the starting position and mode index of all matches. For multi-enzyme Hi-C data, the AC automaton algorithm can identify all relevant enzyme cutting sites (i.e., multiple Motifs) at one time without performing a separate matching operation for each mode. The AC automaton algorithm has an advantage in processing multi-enzyme Hi-C data due to its multi-mode matching ability, high efficiency, and scalability.

[0015] In some embodiments, the Bowtie2 alignment software is used for data alignment in step S3. Thus, by selecting part of the Hi-C data for alignment performance testing, it is found that although Bowtie2 takes longer than BWA to build a reference genome index, the overall running time of Bowtie2 in sequence alignment is generally shorter than BWA, especially when processing large-scale data sets, the speed advantage of Bowtie2 is more obvious. Bowtie2 has become the preferred sequence alignment tool in the Motif-Hi-C analysis method, which can optimize the data processing process and improve the overall analysis efficiency.

[0016] According to a second aspect of the present invention, a method for evaluating Hi-C data quality based on proximity-linked Motif sequences, Motif-Hi-C, is provided for use in Hi-C high-throughput sequencing data analysis. Thus, the application of Motif-Hi-C provides new ideas and solutions for high-throughput sequencing data analysis, and is particularly suitable for large-scale Hi-C data quality analysis, which can greatly improve analysis efficiency and speed.

[0017] According to a third aspect of the present invention, a Hi-C high-throughput sequencing data analysis method is provided, the method comprising the following steps: S1: The Hi-C data quality assessment method Motif-Hi-C based on proximity-linked Motif sequences is used to perform a preliminary assessment of the Hi-C high-throughput sequencing data; S2: According to the results of the preliminary evaluation in step S1, if the evaluation results can be used for subsequent analysis, the matched.fastq file data is compared, filtered and deduplicated to obtain specific valid interaction reads information; S3: Perform downstream analysis on the matched.fastq file data processed in step S2 and the unmatched.fastq file data after noise reduction processing separately or combine them before performing downstream analysis.

[0018] Therefore, after quickly judging the quality of Hi-C data, if the analysis results of the Motif-Hi-C method determine that the data set can continue downstream analysis, there is no need to reanalyze all the data, but only to complete the quality control analysis of the matched.fastq file data. The matched.fastq file data and the unmatched.fastq file data can be analyzed separately for downstream analysis, or they can be combined for downstream analysis. This not only retains the advantage of rapid assessment of data set quality, but also expands the possibility of in-depth analysis of the data, providing a flexible solution for efficient and accurate Hi-C data processing and analysis.

[0019] According to a fourth aspect of the present invention, a method for evaluating Hi-C data quality based on proximity-linked Motif sequences, Motif-Hi-C, is provided for use in evaluating Hi-C data quality. Thus, by adopting this method, the efficiency of Hi-C data quality evaluation can be improved, and the quality of Hi-C data can be evaluated quickly and accurately.

[0020] According to a fifth aspect of the present invention, a method for rapid analysis of large-scale Hi-C high-throughput sequencing data is provided, the method comprising the following steps: S1: Hi-C data storage fastq file for Motif matching: For single enzyme digestion Hi-C data, the KMP algorithm is used to match the Motif, and for multiple enzyme digestion Hi-C data, the AC automaton algorithm is used to match the Motif; S2: Classify the original Hi-C data storage fastq files: Classify the original Hi-C data storage fastq files into matched.fastq and unmatched.fastq files according to the restriction endonuclease cutting sites and Motif characteristics; S3: Perform downstream analysis only on the matched.fastq data in step S2 after quality control.

[0021] Therefore, by classifying large-scale raw data and performing downstream analysis only on matched.fastq data after quality control, not only can effective analysis results be obtained, but the volume of data quality control processing and analysis is also greatly reduced, greatly improving the efficiency of Hi-C data analysis.

[0022] In some embodiments, the method for classifying the original Hi-C data storage fastq file in step S2 comprises the following steps: a1: Screening read IDs containing specific motifs: Read the fastq files containing Hi-C sequencing reads. For each read in the R1.fastq and R2.fastq files, the program will search for read IDs containing specific motifs in their IDs; a2: Output qualified read IDs: For reads whose IDs contain specific motifs, the program will output their IDs into two new files respectively; a3: Merge and remove duplicate IDs: The program will merge the two folders in step a2 into a new file and rename it to ensure that all qualified read IDs are integrated. The merged ID list will then be deduplicated and saved as a new file. a4: Classify fastq files according to the ID list: Use the ID list of the file after deduplication in step a3 to classify the original R1.fastq and R2.fastq files. If the read ID exists in the file after deduplication in step a3, the read is classified as matched, and R1.matched.fastq and R2.matched.fastq files are generated for R1.fastq and R2.fastq, respectively; if the read ID does not exist in the file after deduplication in step a3, the read is classified as unmatched, and R1.unmatched.fastq and R2.unmatched.fastq files are generated for R1.fastq and R2.fastq, respectively.

[0023] Beneficial effects of the present invention: The present invention discloses a Hi-C data quality assessment method Motif-Hi-C based on proximity connection Motif sequences. After the method classifies the Hi-C data, different strategies are used for subsequent processing of different types of data, which can greatly improve the efficiency of Hi-C data quality assessment, and the method is efficient, scientific and accurate. The application of Motif-Hi-C provides new ideas and solutions for high-throughput Hi-C sequencing data analysis, and is particularly suitable for large-scale Hi-C data quality analysis, which can greatly improve analysis efficiency and analysis speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Form a schematic diagram for a specific Motif; Figure 2 Functional flow chart for the Motif-Hi-C method to quickly judge the quality of Hi-C data; Figure 3 This is the percentage of total data noise in matched.fastq and unmatched.fastq files; Figure 4 Statistical graph of the running time of the Motif-Hi-C method and four Hi-C data analysis software (small Hi-C data set); Figure 5 Statistical graph of the running time of the Motif-Hi-C method and four Hi-C data analysis software (large Hi-C dataset); Figure 6 Statistical graphs of the analysis results of the Motif-Hi-C method and four Hi-C data analysis software; Figure 7 Trend statistics of the analysis results of the Motif-Hi-C method and four Hi-C data analysis software; Figure 8A flow chart of the complete upstream analysis (including supplementary process design) of the Motif-Hi-C method; Fig. 9 Statistical graph of data analysis results for matched.fastq and unmatched.fastq files; Fig.10 The data visualization result diagram of matched.fastq and unmatched.fastq files in the C2C12_4_i_M dataset: the red area in the figure represents the interaction between chromatin, the yellow boxes represent TADs (only one of the boxes is indicated by a yellow arrow, but it does not mean that there is only one), and the black dots represent loops (only one of the black dots is indicated by a black arrow, but it does not mean that there is only one); Fig.11 The data visualization result diagram of matched.fastq and unmatched.fastq files in the 293T_2_i_M dataset: the red area in the figure represents the interaction between chromatin, the yellow boxes represent TADs (only one of the boxes is indicated by a yellow arrow, but it does not mean that there is only one), and the black dots represent loops (only one of the black dots is indicated by a black arrow, but it does not mean that there is only one); Fig.12 This is the data visualization result diagram of the matched.fastq and unmatched.fastq files in the PC-3_4_i_M dataset: the red area in the figure represents the interaction between chromatin, the yellow boxes represent TADs (only one of the boxes is indicated by a yellow arrow, but it does not mean that there is only one), and the black dots represent loops (only one of the black dots is indicated by a black arrow, but it does not mean that there is only one). DETAILED DESCRIPTION

[0025] Example 1 Design and process of Motif-Hi-C method.

[0026] Based on the analysis process of existing Hi-C data analysis software, most of the time is spent on analysis and comparison in the upstream stage. Therefore, how to shorten the time required for analysis and comparison in the upstream stage will be the key to improving the efficiency of Hi-C data quality control.

[0027] When performing Hi-C experiments, restriction endonucleases are used to identify and cut specific restriction sites on DNA fragments. Then, after biotin filling and ligation steps, the originally non-adjacent DNA fragments are connected together to form a unique connection sequence Motif. Since each restriction endonuclease has its own specific restriction site, the chimeric short sequence (Motif sequence) formed by adjacent ligation obtained by biotin filling of the DNA fragments after enzyme cleavage is also specific. Therefore, the effectively interacting DNA fragments obtained must contain this specific Motif (such as Figure 1 As shown in the figure, taking HpyCH4IV endonuclease as an example, the enzyme recognizes ACGT as the restriction site, and the formed motif is ACGCGT. All DNA fragments that are cut by HpyCH4IV and have effective interactions must contain this specific motif).

[0028] In combination with the Hi-C sequencing data type, the present invention proposes a Hi-C data quality assessment method based on neighboring connection Motif sequences (hereinafter referred to as the "Motif-Hi-C" method). This method classifies the original Hi-C data storage fastq files into matched.fastq (containing Motif sequences) and unmatched.fastq (not containing Motif sequences) files according to Motif. Then, each part of the classified file data is processed separately according to its data characteristics to reduce the amount of data required for upstream comparison analysis, improve data analysis efficiency and speed, and achieve the purpose of quickly judging the quality of Hi-C data. Figure 2 This is a flow chart of the method, which mainly includes: matching and classifying Motifs of original fastq files, comparing, filtering, and deduplicating unmatched.fastq file data, and merging and statistics of analysis results.

[0029] 1.1 Hi-C data storage fastq file for Motif matching.

[0030] For single-enzyme Hi-C data, the KMP algorithm is used to match the Motif, and for multiple-enzyme Hi-C data, the AC automaton algorithm is used to match the Motif.

[0031] (1) Matching strategy of single enzyme digestion Hi-C data based on KMP algorithm: KMP algorithm is a string matching algorithm used to find the occurrence position of a pattern string in a main text string. It is an efficient linear time complexity algorithm. The classification strategy of single enzyme digestion Hi-C data based on KMP algorithm is mainly used to improve the efficiency of finding Motif. This algorithm makes full use of the longest common prefix and suffix length before each position of Motif, that is, the next array is used to realize the movement and backtracking of Motif, while the read sequence does not backtrack, so as to reduce the number of backtracking when the matching fails, achieve the maximum movement amount, and thus improve the matching efficiency. Its time complexity is , where N is the read sequence length and M is the motif length. The execution flow of this classification strategy is as follows: 1) Build a partial match table based on Motif to quickly move the pointer during the matching process. And initialize a partial match table with the same length as Motif, with all initial values ​​0.

[0032] 2) Starting from the second character of the Motif, calculate the longest common prefix and suffix length of each position in turn and record them in the partial match table. During the matching process, the constructed partial match table is used to search for the Motif in the read sequence.

[0033] 3) Initialize two pointers to point to the start position of the reads sequence and Motif respectively.

[0034] 4) Move the pointers in the reads sequence one by one, compare the base characters at the corresponding positions, and adjust the position of the Motif according to the partial match table. If the current base character matches successfully, both pointers move back one position at the same time. If the current base character does not match, and the value of the partial match table is greater than 0, move the Motif to the right (skip the matched part) and keep the pointer of the reads sequence unchanged. If the current base character does not match, and the value of the partial match table is 0, move both the Motif and reads sequence pointers to the right one position.

[0035] 5) Repeat the above process until the match is completed or the pointer of the reads sequence moves to the end.

[0036] (2) Classification strategy of multi-enzyme digestion Hi-C data based on AC automaton algorithm: AC automaton algorithm is a multi-mode string matching algorithm, which is widely used in the field of multi-mode matching, such as banned word search and replacement, search keyword search, etc. Its time complexity is , where N is the read sequence length, M is the total length of all Motifs, and Z is the number of matching results. The execution flow of this classification strategy is as follows: 1) Construct a Trie tree as the search data structure of the AC automaton.

[0037] 2) Construct a fail pointer so that when the current base character fails to match, it jumps to the base character with the longest common prefix and suffix to continue matching. Just like the KMP algorithm, when the AC automaton fails to match the current character, it uses the fail pointer to jump.

[0038] 3) Scan the reads sequence for matching.

[0039] 1.2 Classify the original Hi-C data storage fastq files.

[0040] The fastq file obtained after the Hi-C library is sequenced by the second generation sequencing is a four-line unit, which represents the information of the sequencing data. The contents of the four lines are as follows: the first line is the ID of the reads, the second line is the sequence information of the reads, the third line is the "+" separator, and the fourth line is the sequence quality of the reads.

[0041] According to the restriction endonuclease cutting site and Motif characteristics, the Hi-C data storage fastq files after Motif matching are classified into matched.fastq (containing Motif sequence) and unmatched.fastq (not containing Motif sequence). The specific classification method is as follows: 1) Screening read IDs containing specific motifs: First, read the fastq files containing Hi-C sequencing reads (i.e., R1.fastq and R2.fastq). For each read in the R1.fastq and R2.fastq files, the program searches for read IDs containing specific motifs (matching the motif sequence) (e.g., ACGCGT) in their IDs (usually located after the @ symbol before each line of read sequence).

[0042] 2) Output qualified read IDs: For reads with specific motifs in their IDs, the program will output their IDs to two new files. Assume that these two files are named R1_ACGCGT_ID.txt and R2_ACGCGT_ID.txt. In this way, R1_ACGCGT_ID.txt contains all qualified read IDs in R1.fastq, while R2_ACGCGT_ID.txt contains the corresponding IDs in R2.fastq.

[0043] 3) Merge and remove duplicate IDs: The program will merge the IDs in R1_ACGCGT_ID.txt and R2_ACGCGT_ID.txt into a new file, such as ACGCGT_combined_ID.txt, to ensure that all qualified readIDs are integrated. The merged ID list will then be deduplicated and saved as a new file, such as ACGCGT_unique_ID.txt.

[0044] 4) Classify fastq files based on the ID list: Use the ID list in unique_ID.txt to classify the original R1.fastq and R2.fastq files. If the read's ID exists in unique_ID.txt, the read (including its sequence and quality information) is classified as "matched". For R1.fastq and R2.fastq, R1.matched.fastq and R2.matched.fastq files are generated, respectively. Similarly, if the read's ID does not exist in unique_ID.txt, the read is classified as "unmatched". For R1.fastq and R2.fastq, R1.unmatched.fastq and R2.unmatched.fastq files are generated, respectively.

[0045] Subsequently, the matched.fastq and unmatched.fastq files were processed separately.

[0046] 1.3 Selection of software for paired-end sequence alignment.

[0047] In the Hi-C data analysis process, sequence alignment is a key and most time-consuming step, so choosing an efficient alignment software is crucial to optimizing the time efficiency of the entire analysis process. Currently, the widely used alignment software mainly includes BWA and Bowtie2, both of which are implemented based on the BWT algorithm. In order to evaluate the performance of these two software and determine the alignment tool that best suits the data analysis needs of the present invention, the present invention conducted a series of performance tests. These tests were performed in a twenty-thread environment. Specifically, in terms of alignment mode settings, BWA uses the mem mode, while Bowtie2 uses the --end-to-end mode for single-end alignment, and the remaining parameter settings follow the default configuration of each mode. This performance evaluation covers two main aspects: first, the time efficiency of the two in building a reference genome index is compared; second, their running time when performing sequence alignment tasks is evaluated. Revealed by the final test results, it was found that in the index building step, Bowtie2 takes longer than BWA, while for most data, in terms of sequence alignment, Bowtie2 has a shorter running time than BWA. But overall, Bowtie2 is still better than BWA in terms of running speed. And, through the final alignment results, it is found that although BWA is slightly better than Bowtie2, the difference between the two is not large. Finally, considering the performance of running speed and alignment rate, Bowtie2 was finally selected as the sequence alignment tool in the Motif-Hi-C analysis method to optimize the data processing flow of Motif-Hi-C and improve the efficiency of the overall analysis.

[0048] 1.4 Processing strategy for matched.fastq and unmatched.fastq file data.

[0049] In this paper, 12 sets of Hi-C datasets were selected for analysis, covering different cell types and restriction enzymes commonly used in Hi-C research (all 12 sets of Hi-C datasets were provided by the Institute of Agricultural Genomics, Chinese Academy of Agricultural Sciences). The above datasets are divided into two categories: 5 small Hi-C datasets (C2C12_1_D_S, 293T_1_i_S, PC-3_1_i_S, PC-3_2_i_S, PC-3_3_i_S) and 7 large Hi-C datasets (C2C12_2_D_M, C2C12_3_D_M, C2C12_4_i_M, 293T_2_i_M, PC-3_4_i_M, PC-3_5_i_M, PC-3_6_i_M), and the detailed information is shown in Table 1. The relevant Hi-C data have been uploaded to the NCBI database (URL: https: / / www.ncbi.nlm.nih.gov / ). The original data can be obtained by entering the project numbers PRJNA1203007 and PRJNA1201070 (PRJNA1203007 contains mouse Hi-C library data, namely C2C12_1_D_S, C2C12_2_D_M, C2C12_3 _D_M, C2C12_4_i_M library data; PRJNA1201070 contains human Hi-C library data, namely 293T_1_i_S, 293T_2_i_M, PC-3_1_i_S, PC-3_2_i_S, PC-3_3_i_S, 293T_2_i_M, PC-3_4_i_M, PC-3_5_i_M, PC-3_6_i_M library data).

[0050] Table 1 Detailed information of 12 Hi-C datasets Dataset name Cell Type Fragment length (bp) Connection system Motif Sequence C2C12_1_D_S C2C12 50-1000 dilution ligation(Mouse_C2C12_DD) 5'-GATCCGT-3'5'-ACGGATC-3'5'-GATCGATC-3'5'-ACGCGT-3' 293T_1_i_S HEK 293T 80-680 in situ ligation(Human_293T_DD) 5'-GATCCGT-3'5'-ACGGATC-3'5'-GATCGATC-3'5'-ACGCGT-3' PC-3_1_i_S Human_PC-3 150-700 in situ ligation(Human_PC-3_i1_SD) 5'-GATCGATC-3' PC-3_2_i_S Human_PC-3 150-700 in situ ligation(Human_PC-3_i2_SD) 5'-GATCGATC-3' PC-3_3_i_S Human_PC-3 150-700 in situ ligation(Human_PC-3_c2_SD) 5'-GATCGATC-3' C2C12_3_D_M C2C12 50-1000 dilution ligation(Mouse_C2C12_DD) 5'-GATCCGT-3'5'-ACGGATC-3'5'-GATCGATC-3'5'-ACGCGT-3' C2C12_4_i_M C2C12 50-1000 in situ ligation(Mouse_C2C12_DD) 5'-GATCGATC-3' C2C12_2_D_M C2C12 50-700 dilution ligation(Mouse_C2C12_SD) 5'-GATCCGT-3'5'-ACGGATC-3'5'-GATCGATC-3'5'-ACGCGT-3' 293T_2_i_M HEK 293T 80-680 in situ ligation(Human_293T_DD) 5'-GATCCGT-3'5'-ACGGATC-3'5'-GATCGATC-3'5'-ACGCGT-3' PC-3_4_i_M Human_PC-3 50-900 in situ ligation(Human_PC-3_c1_SD) 5'-GATCGATC-3' PC-3_5_i_M Human_PC-3 50-900 in situ ligation(Human_PC-3_i1_1_SD) 5'-GATCGATC-3' PC-3_6_i_M Human_PC-3 50-900 in situ ligation(Human_PC-3_i1_2_SD) 5’-GATCGATC-3’ (1) Processing strategy for unmatched.fastq file data (noise reduction): For reads in unmatched.fastq files, given that they contain a lot of data noise, the classic Hi-C Pro (Servant et al., 2015, Genome Biology) Hi-C data processing process with the Bowtie2 alignment process was used to perform alignment, filtering, and deduplication steps. The PCR amplification deduplication multiple was calculated based on the deduplication of the unmatched.fastq file data.

[0051] Processing strategy for matched.fastq file data (simulated deduplication): In the Motif-Hi-C method, no real alignment, filtering and deduplication steps are performed on the matched.fastq file data. Instead, the PCR amplification deduplication multiple calculated from the unmatched.fastq file data is used to simulate PCR amplification deduplication on the matched.fastq file evaluation results (specifically: when calculating various evaluation values ​​in the matched.fastq file data, the final matched.fastq evaluation value is obtained by directly dividing the fast evaluation value by the PCR amplification deduplication multiple. The PCR amplification deduplication multiple is obtained based on the real PCR amplification deduplication multiple calculated after the unmatched.fastq file data is truly deduplicated). Therefore, in order to verify the accuracy of this processing strategy, we first performed a statistical analysis on the total noise ratio of the matched.fastq and unmatched.fastq file data storing 12 groups of Hi-C data sets, as shown in the figure. Figure 3 As shown, the results show that compared with the reads in the unmatched.fastq file, the reads in the matched.fastq file of the dataset contain very little noise, so the reads in the matched.fastq file have a higher probability of belonging to valid interaction reads. For example, Hi-CPro and HiCUP were used to analyze the matched.fastq and unmatched.fastq file data of the PC-3_1_i_S dataset. The results show that in the reads data of the matched.fastq file, the total number of noise is 137,476 (pairs), accounting for 8.55% of the total noise of the dataset. In the reads data of the unmatched.fastq file, the total number of noise increased significantly to 1,470,858 (pairs), accounting for 91.45% of the total noise of the dataset. Therefore, in order to quickly evaluate the quality of Hi-C data and reduce the time consumed by operations such as alignment, filtering, and deduplication, for the data processing of matched.fastq files, Motif-Hi-C rapid evaluation can be performed without filtering its noise, and the reads obtained after classification are regarded as the valid interaction number (validinteraction rmdup).

[0052] After completing the above steps, the data processing and evaluation results of the matched.fastq and unmatched.fastq files are merged for comprehensive analysis and the ratio of the valid interaction number (valid interaction rmdup) to the total read pairs (total readpairs) is calculated. This ratio is an important indicator for evaluating the overall quality of the Hi-C dataset, enabling researchers to quickly make a preliminary assessment of the quality of the Hi-C dataset. Through this method, although the accuracy of the matched.fastq file data may be slightly affected by the lack of quality control, this method is still within an acceptable error range, which significantly saves analysis time. This enables researchers to quickly determine whether the dataset is suitable for further downstream stage analysis, thereby improving analysis efficiency while ensuring data quality.

[0053] Example 2 Feasibility verification of the Motif-Hi-C method classification strategy.

[0054] Although the study in Example 1 shows that the noise level of the matched.fastq file data is extremely low, in order to further verify the feasibility of the Motif-Hi-C method for the fastq file classification strategy, the present invention adopts two evaluation methods. First, considering that the classification strategy of the Motif-Hi-C method does not further subdivide the reads containing Motif according to the number of Motifs, it may affect the accuracy of the analysis results. Therefore, the present invention evaluates the number of reads containing two or more Motifs to verify the rigor of the classification strategy of this method. Secondly, the present invention compares the repeatability of the effective interaction reads obtained by the classification strategy with the reads obtained after quality control analysis using conventional analysis software. Here, the present invention uses the HiCUP tool. In this way, the rationality and accuracy of the method are evaluated.

[0055] 2.1 Evaluate the number of reads containing multiple motifs.

[0056] In the present invention, although according to the principle of Hi-C experiment, it is known that the probability of generating reads containing multiple Motifs in conventional Hi-C experiments is low. However, in order to ensure the rigor of the experiment, the present invention analyzed 3 groups of Hi-C data sets to calculate the number of reads containing two or more Motifs in these data. Taking the PC-3_1_i_S data set as an example, the proportion of reads with different numbers of Motifs to the total number of reads is discussed here to determine the impact of each classification on the overall data. The results show that in the R1.fastq file, the number of reads containing one Motif is 2,649,049, the number of reads containing two Motifs is 1,646, and the number of reads containing two or more Motifs is only 3. Among them, the number of reads containing two or more Motifs accounts for 0.06% of the total number of reads. Similarly, in the R2.fastq file, the number of reads containing one motif is 2,104,521, the number of reads containing two motifs is 1,337, and no reads containing more than two motifs were found. The number of reads containing two motifs accounts for 0.06% of the total number of reads.

[0057] On the other hand, the type of Hi-C sequencing data takes into account the situation where read1 and read2 just cover the sequencing fragment. In this case, the Motif is just divided equally by read1 and read2, and the Motif cannot be directly detected in read1 or read2. Therefore, the present invention further analyzes the connection between the reverse complementary sequence of the reads in R2.fastq and the corresponding reads sequence in R1.fastq, and finds that the number of reads containing one Motif is 4,504,923, the number of reads containing two Motifs is 130,312, and the number of reads containing more than two Motifs is 294. Among them, the number of reads containing two or more Motifs accounts for 2.81% of the total number of reads. Therefore, it can be concluded that the proportion of reads containing two or more Motifs in the overall reads is extremely low, and some invalid reads may be included. This shows that the failure to classify these reads separately has little effect on the accuracy of evaluating the quality of Hi-C data.

[0058] 2.2 Repeatability test of Motif-Hi-C and HiCUP analysis results.

[0059] In order to evaluate the accuracy of the valid reads retained by the Motif-Hi-C method during the analysis process, the present invention compares the repetitiveness of the valid reads retained by the Motif-Hi-C and HiCUP methods. Specifically, the present invention compares the IDs of the reads finally retained by the two methods, aiming to evaluate the proportion of the same reads retained by the two methods in the reads after quality control of the two methods, thereby verifying the accuracy of Motif-Hi-C. It should be pointed out that in the processing strategy of the matched.fastq file data, the Motif-Hi-C method only simulates the deduplication processing based on the unmatched.fastq file data and does not perform the actual deduplication step. Therefore, when performing repeatability detection with the HiCUP software, the ID compared is the reads ID that has been filtered but not deduplicated. The results show that in most Hi-C data sets, the reads retained by the Motif-Hi-C method almost include the reads retained after filtering by the HiCUP method, and the reads repetition with the HiCUP control group is at least 75%. This shows that the valid interaction reads retained by the Motif-Hi-C method are real and valid in most cases.

[0060] Example 3 Using Motif-Hi-C to quickly determine the quality of Hi-C data.

[0061] After feasibility verification, actual data testing was performed to evaluate the performance of the Motif-Hi-C method in terms of rapid judgment of Hi-C data quality and reliability of its analysis results. By comparing the analysis results of the Motif-Hi-C method with those of four mainstream Hi-C data analysis software, this example reveals the advantage of Motif-Hi-C in terms of analysis speed.

[0062] First, in order to verify the advantage of Motif-Hi-C in analysis speed, the present invention selected 5 small Hi-C datasets (C2C12_1_D_S, 293T_1_i_S, PC-3_1_i_S, PC-3_2_i_S, PC-3_3_i_S) and 5 large Hi-C datasets (C2C12_2_D_M, C2C12_3_D_M, C2C12_4_i_M, 293T_2_i_M, PC-3_5_i_M) from the above 12 Hi-C datasets according to the cell type and the type of restriction endonuclease used. And in order to ensure the consistency of the experiment, the present invention sets the number of task threads of the small Hi-C dataset to two threads, the number of task threads of the large Hi-C dataset to twenty threads, and calculates the running time of each dataset. Then, the present invention compared the performance of Motif-Hi-C with the four Hi-C data analysis software used by the above-mentioned research institutes in terms of running time. Figure 4 , Figure 5 As shown in the figure, after analyzing 10 sets of Hi-C data sets, the research results show that the Motif-Hi-C method used in the present invention is significantly better than the other four Hi-C data analysis software in terms of running time. For example, for the 293T_1_i_S data set, the analysis time of Motif-Hi-C is almost the shortest among the four Hi-C data analysis software, about half of the Juicer software. Moreover, compared with Hi-C Pro and HiCUP, two softwares with higher reliability in analysis results, Motif-Hi-C also significantly reduced the time required for data analysis and quality control in the test of 10 sets of Hi-C data sets. For example, for the PC-3_1_i_SHi-C data set, Motif-Hi-C only takes 9,589 seconds to evaluate the data quality, while Hi-C Pro and HiCUP take 26,389 seconds and 17,596 seconds respectively. It is worth noting that this time efficiency advantage will be more obvious in data sets with high motif ratios, such as 293T_2_i_M, C2C12_2_D_M, C2C12_4_i_M, etc. This proves the unique advantage of the Motif-Hi-C method in the study of rapid judgment of Hi-C data quality, and further emphasizes the application potential of Motif-Hi-C in the efficient processing of Hi-C data containing specific motifs.

[0063] Furthermore, in order to verify the reliability of the Motif-Hi-C analysis results, the Hi-C data quality was judged by valid interaction rmdup / total read pairs. Therefore, 12 sets of Hi-C data sets were used to compare the performance of the Motif-Hi-C method and four Hi-C data analysis software in terms of valid interaction rmdup / total read pairs. The results showed that Figure 6 It can be observed that there is no significant difference between the indicators obtained by the Motif-Hi-C method and the other four Hi-C data analysis software, and these differences are in line with expectations, which further demonstrates that it is feasible to use the Motif-Hi-C method to quickly judge the quality of Hi-C data. In addition, the present invention analyzes the change trend of the Motif-Hi-C method and the four Hi-C data analysis software in the indicator of valid interaction rmdup / total read pairs for 12 groups of Hi-C data. The results are as follows Figure 7 The results show that the trend of Motif-Hi-C analysis results is very close to that of the other four Hi-C data analysis software. This result shows that when it is necessary to screen the highest quality data set from multiple sets of data, the Motif-Hi-C method can provide a fast and relatively accurate choice.

[0064] Example 4 Hi-C high-throughput sequencing data analysis method (supplementary process based on Motif-Hi-C analysis).

[0065] Although the Motif-Hi-C method for quickly judging the quality of Hi-C data mentioned above can effectively realize a quick preliminary assessment of the quality of the Hi-C data set, since the alignment step is not performed on the reads data of the matched.fastq file, the exact location information of this part of the reads on the reference genome has not yet been determined, and therefore it cannot be directly used for downstream stage analysis. In response to this limitation, the present invention proposes a supplementary process: performing alignment, filtering and deduplication operations on the matched.fastq file data to obtain specific effective interaction reads information. In this way, after the researcher quickly judges the quality of the Hi-C data, if it is determined that the set of data sets can continue the downstream stage analysis based on the analysis results of the Motif-Hi-C method, there is no need to re-analyze all the data, but only needs to complete the quality control analysis of the matched.fastq file data, such as Figure 8As shown, the specific steps include: S1: Preliminary evaluation of Hi-C high-throughput sequencing data based on the Motif-Hi-C analysis method; S2: According to the result of the preliminary evaluation in step S1, if the evaluation result can be used for subsequent analysis, the matched.fastq file data is compared, filtered and deduplicated to obtain specific effective interaction reads information; S3: The matched.fastq file data processed in step S2 and the unmatched.fastq file data after noise reduction are separately analyzed for downstream analysis or merged for downstream analysis.

[0066] In addition, in order to explore the potential association between Motif and different chromatin structures. After classifying the initial fastq files, the present invention compares, filters and removes duplicates of the matched.fastq and unmatched.fastq file data to obtain two independent analysis results. In this way, researchers can perform independent downstream analysis on the matched.fastq and unmatched.fastq data of the same Hi-C data set to analyze the relationship between the distribution of Motif and chromatin structure. Finally, the Motif-Hi-C method can also use the Samtools tool to merge the two parts of the results to provide complete upstream analysis data for further analysis. Such a method not only retains the advantage of rapid assessment of data set quality, but also expands the possibility of in-depth analysis of data, providing a flexible solution for efficient and accurate Hi-C data processing and analysis.

[0067] Example 5 Motif-Hi-C was used to analyze the impact of Motif distribution on Hi-C data.

[0068] After the Motif-Hi-C method was used to quickly evaluate the quality of the Hi-C data set, this embodiment further studied the impact of whether the Motif was contained on the Hi-C data under the Motif-Hi-C classification strategy, that is, analyzing the difference between the matched.fastq and unmatched.fastq file data. On the other hand, considering that the chromatin structure information contained in the large Hi-C data set is relatively rich, it is more conducive to the comparison of subsequent downstream analysis. Therefore, the present invention selected 7 groups of large Hi-C data sets (C2C12_2_D_M, C2C12_3_D_M, C2C12_4_i_M, 293T_2_i_M, PC-3_4_i_M, PC-3_5_i_M, PC-3_6_i_M), and used Motif-Hi-C to successfully obtain the matched.fastq part and the unmatched.fastq part of each group of Hi-C data sets. In addition, if an in-depth analysis is to be performed, it is necessary to understand the specific alignment position of the data on the reference genome and its alignment quality and other information. Therefore, the present invention uses the supplementary process of the Motif-Hi-C method (Example 4) to further perform upstream analysis steps such as alignment, filtering, and deduplication on the matched.fastq and unmatched.fastq file data.

[0069] Through this series of processing, the present invention generates two quality-controlled BAM files for the matched.fastq and unmatched.fastq parts, respectively. Further, the present invention uses the HiCUP2Juicer script to convert the two BAM result files of the matched.fastq part and the unmatched.fastq part into files that can be used for downstream analysis.

[0070] 5.1 Quality control analysis and evaluation of the impact of motif distribution on Hi-C data.

[0071] By adopting the Motif-Hi-C analysis method, the present invention performed a comprehensive and detailed quality control analysis and evaluation on the matched.fastq and unmatched.fastq file data in 7 large Hi-C data sets previously studied. The detailed analysis results are shown in Tables 2-8.

[0072] Table 2 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of PC-3_5_i_M dataset Analysis indicators matched.fastq unmatched.fastq Total_read pairs_processed 150,696,704 175,001,088 Unique_paired_alignments 90,782,504 84,799,630 Mapping_rate(%) 60.2 48.5 Valid_interaction_pairs 89,208,249 71,262,153 Dangling / Internal_end_pairs 5,659 5,762,524 Religation_pairs 499,785 3,394,689 Self Cycle pairs 216,703 191,919 Dumped_pairs 315,115 2,739,162 Valid_interaction_rmdup 20,478,527 16,611,070 Trans_interaction 3,420,380 2,821,445 Cis_interaction 17,058,147 13,789,625 Cis_shortRange 2,482,785 1,976,044 Cis_longRange 14,575,362 11,813,581 Valid_interaction_rmdup / Total_read pairs(%) 13.6 9.5 Table 3 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of PC-3_6_i_M dataset Analysis indicators matched.fastq unmatched.fastq Total_read pairs_processed 170,133,672 184,531,187 Unique_paired_alignments 104,717,823 91,679,225 Mapping_rate(%) 61.6 49.7 Valid_interaction_pairs 102,898,627 77,522,777 Dangling / Internal_end_pairs 6,750 6,093,613 Religation_pairs 576,968 3,549,110 Self Cycle pairs 250,712 204,389 Dumped_pairs 366,406 2,812,044 Valid_interaction_rmdup 32,647,838 25,186,139 Trans_interaction 5,456,118 4,261,023 Cis_interaction 27,191,720 20,925,116 Cis_shortRange 3,966,399 3,002,199 Cis_longRange 23,225,321 17,922,917 Valid_interaction_rmdup / Total_read pairs(%) 19.2 13.6 Table 4 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of PC-3_4_i_M dataset Analysis indicators matched.fastq unmatched.fastq Total_read pairs_processed 109,219,079 151,511,239 Unique_paired_alignments 73,417,531 67,998,606 Mapping_rate(%) 67.2 44.9 Valid_interaction_pairs 71,780,966 41,754,798 Dangling / Internal_end_pairs 6,927 12,790,722 Religation_pairs 543,246 6,502,823 Self Cycle pairs 178,686 105,642 Dumped_pairs 328,135 4,425,571 Valid_interaction_rmdup 50,060,865 28,979,624 Trans_interaction 7,930,430 5,024,262 Cis_interaction 42,130,435 23,955,362 Cis_shortRange 6,684,073 3,738,807 Cis_longRange 35,446,362 20,216,555 Valid_interaction_rmdup / Total_read pairs(%) 45.8 19.1 Table 5 Quality control analysis and evaluation results of matched.fastq and unmatched.fastq of 293T_2_i_M dataset Analysis metrics matched.fastq unmatched.fastq Total_read pairs_processed 262,067,352 130,196,103 Unique_paired_alignments 191,992,047 96,045,752 Mapping_rate(%) 73.3 73.8 Valid_interaction_pairs 185,856,147 70,049,775 Dangling / Internal_end_pairs 20,425 9,174,466 Religation_pairs 737,638 9,915,890 Self Cycle pairs 143,699 65,878 Dumped_pairs 375,121 862,802 Valid_interaction_rmdup 133,623,495 50,570,505 Trans_interaction 20,526,884 7,653,309 Cis_interaction 113,096,611 42,917,196 Cis_shortRange 15,469,569 5,843,044 Cis_longRange 97,627,042 37,074,152 Valid_interaction_rmdup / Total_read pairs(%) 51.0 38.8 Table 6 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of C2C12_3_D_M dataset Analysis metrics matched.fastq unmatched.fastq Total_read pairs_processed 99,443,969 110,724,395 Unique_paired_alignments 57,403,247 52,471,550 Mapping_rate(%) 57.7 47.4 Valid_interaction_pairs 56,262,780 37,551,549 Dangling / Internal_end_pairs 14,561 3,414,703 Religation_pairs 158,418 5,525,878 Self Cycle pairs 142,627 125,974 Dumped_pairs 628,568 627,833 Valid_interaction_rmdup 39,973,160 26,264,916 Trans_interaction 21,591,017 14,825,196 Cis_interaction 18,382,143 11,439,720 Cis_shortRange 729,277 473,954 Cis_longRange 17,652,866 10,965,766 Valid_interaction_rmdup / Total_read pairs(%) 40.2 23.7 Table 7 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of C2C12_2_D_M dataset Analysis metrics matched.fastq unmatched.fastq Total_read pairs_processed 334,641,526 156,490,829 Unique_paired_alignments 199,701,363 92,247,230 Mapping_rate(%) 59.7 58.9 Valid_interaction_pairs 195,058,882 48,702,706 Dangling / Internal_end_pairs 33,534 6,517,301 Religation_pairs 504,085 19,420,256 Self Cycle pairs 233,502 84,980 Dumped_pairs 3,415,916 1,907,807 Valid_interaction_rmdup 18,083,463 6,197,428 Trans_interaction 9,852,420 3,297,341 Cis_interaction 8,231,043 2,900,087 Cis_shortRange 302,976 203,072 Cis_longRange 7,928,067 2,697,015 Valid_interaction_rmdup / Total_read pairs(%) 5.4 4.0 Table 8 Statistics of quality control analysis and evaluation results of matched.fastq and unmatched.fastq of C2C12_4_i_M dataset Analysis metrics matched.fastq unmatched.fastq Total_read pairs_processed 92,786,532 59,894,785 Unique_paired_alignments 66,054,238 44,122,260 Mapping_rate(%) 71.2 73.7 Valid_interaction_pairs 64,762,493 41,818,893 Dangling / Internal_end_pairs 5,348 409,842 Religation_pairs 326,079 859,329 Self Cycle pairs 82,559 59,220 Dumped_pairs 462,292 248,084 Valid_interaction_rmdup 35,558,112 23,833,974 Trans_interaction 4,062,338 2,614,296 Cis_interaction 31,495,774 21,219,678 Cis_shortRange 3,556,681 2,408,887 Cis_longRange 27,939,093 18,810,791 Valid_interaction_rmdup / Total_read pairs(%) 23.6 13.6 The analysis results in Table 2-8 show that the reads of the matched.fastq file contain more valid interactions than the reads of the unmatched.fastq file. Specifically, when analyzing the interaction types in detail, it was found that the number of interactions in the matched.fastq part containing the motif was significantly higher than that in the unmatched.fastq part without the motif, whether it was transition or cis interaction. Especially in the cis interaction type, the matched.fastq file data detected more abundant interactions. In addition, the indicator of valid interaction rmdup / total read pairs in the matched.fastq file is also significantly higher than that in the unmatched.fastq part (such as Figure 9). This indicates that the matched.fastq part containing the motif has higher data quality than the unmatched.fastq part without the motif. On the other hand, in terms of Hi-C data noise, the overall noise level of the matched.fastq part was observed to be lower. However, in the self cycle pairs type of noise, although its number is very low in both types of data, the matched.fastq part is slightly higher than the unmatched.fastq part. This phenomenon may be due to the fact that a motif is generated during the formation of self cycle pairs, and this motif, like the motif contained in the valid DNA fragment, will also be recognized by the Motif-Hi-C method, which is consistent with the experimental principle.

[0073] By deeply analyzing the quality control results of matched and unmatched.fastq file data, we can better understand the relationship between Motif and data quality and interaction characteristics, and provide a solid foundation for subsequent chromatin structure analysis and functional research.

[0074] 5.2 Analysis of chromatin structure based on the impact of motif distribution on Hi-C data.

[0075] After completing the quality control analysis, in order to explore whether there is a potential correlation between these motifs and the chromatin structure at each level, downstream chromatin structure analysis was performed on the matched.fastq and unmatched.fastq file data. Through detailed analysis, the study observed significant differences between the two parts of the data. The matched.fastq part containing the motif carried more abundant chromatin structure information, as shown in Tables 9 and 10. Whether in the detection of loops or TADs, the matched.fastq part significantly exceeded the unmatched.fastq part. Taking the 293T_2_i_M dataset as an example, 915 loops and 758 TADs were detected in the unmatched.fastq part, while 3,208 loops and 2,583 TADs were detected in the matched.fastq part. Even in the C2C12_3_D_M and C2C12_2_D_M datasets with poor data quality, it was found that the vast majority of chromatin structure information was concentrated in the matched.fastq part. In addition, by comparing the two compartments of the data, the study found that the values ​​of the first principal component calculated for the matched.fastq part are generally higher than those calculated for the unmatched.fastq part, which also indicates that the matched.fastq file data contains a stronger compartment signal.

[0076] Table 9 Statistics of TAD numbers detected by Matched.fastq and unmatched.fastq Dataset name (TAD) matched.fastq unmatched.fastq PC-3_5_i_M 67 33 PC-3_6_i_M 247 116 PC-3_4_i_M 666 189 293T_2_i_M 2583 758 C2C12_3_D_M 20 1 C2C12_2_D_M 0 0 C2C12_4_i_M 277 72 Table 10 Statistics of loop numbers detected by Matched.fastq and unmatched.fastq Dataset name (Loop) matched.fastq unmatched.fastq PC-3_5_i_M 409 251 PC-3_6_i_M 1,069 612 PC-3_4_i_M 2,356 960 293T_2_i_M 3,208 915 C2C12_3_D_M 4 1 C2C12_2_D_M 1 0 C2C12_4_i_M 597 243 After completing the downstream chromatin structure analysis, in order to more intuitively see the difference in the number of TADs and loops in the matched.fastq and unmatched.fastq file data, the present invention comprehensively considers the factors of cell type and restriction endonuclease type and takes the 0mb to 90mb region of chromosome 2 of the three groups of data of 293T_2_i_M, PC-3_4_i_M, and C2C12_4_i_M as an example for visualization. The results are as follows Figures 10 - 12 As shown in: Figure 10 The visualization result of C2C12_4_i_M data is shown below. Figure 11 The visualization results of 293T_2_i_M data are shown below. Figure 12 For PC-3_4_i_M data visualization results,Figures 10 - 12 The red area represents the interaction between chromatin, the yellow box represents TADs, and the black dots represent loops. The darker the red, the higher the frequency of interaction within chromatin. Figures 10 - 12 It can be seen that the matched.fastq part of the three sets of data has a higher interaction frequency than the unmatched.fastq part, and contains more TADs and loops. This further proves that the matched.fastq part has richer chromatin interaction information than the unmatched.fastq part.

[0077] In summary, combined with the results of quality control analysis, it is believed that when analyzing Hi-C data and chromatin structure, it is very important to identify and deeply analyze DNA fragments containing Motifs. In addition, the present invention also believes that increasing the proportion of reads containing Motifs can help improve the overall quality of the data set, thereby helping researchers obtain more three-dimensional chromatin structure information. In addition, the research of the present invention also shows that through data classification processing strategies, a new perspective can be provided for analyzing Hi-C data. In particular, when there is still potential for application to large-volume samples in the future, in order to save analysis time and computing resources, it is also feasible to only analyze DNA interaction information containing Motifs (matched.fastq data). If only matched.fastq data is quality controlled and analyzed, not only can effective analysis results (chromatin interaction information) be obtained, but also the data volume of data quality control processing and analysis is greatly reduced, which can greatly improve the efficiency of Hi-C data analysis, and is particularly suitable for large-scale Hi-C high-throughput sequencing data rapid analysis. When large-scale Hi-C high-throughput sequencing data is rapidly analyzed, the method may be: S1: Motif matching of Hi-C data storage fastq files: KMP algorithm is used for Motif matching of single-enzyme Hi-C data, and AC automaton algorithm is used for Motif matching of multi-enzyme Hi-C data (specific matching method can be seen in Example 1); S2: Classification of original Hi-C data storage fastq files: Classification of original Hi-C data storage fastq files into matched.fastq and unmatched.fastq files according to restriction endonuclease cleavage sites and Motif characteristics (specific classification method can be seen in Example 1); S3: Downstream analysis is performed only on the matched.fastq data in step S2 after quality control.

[0078] In summary, the present invention has developed an innovative method, named Motif-Hi-C method. This method is mainly used to quickly evaluate the quality of Hi-C data. The core of this method is to use Motif to effectively classify the original fastq data and then adopt a differentiated processing strategy according to the characteristics of each part of the Hi-C data. Thereby significantly saving the time of comparison, thereby greatly improving the efficiency of upstream data analysis. Furthermore, the feasibility analysis of the classification strategy adopted by Motif-Hi-C was carried out to ensure its scientificity and rationality. Then, the actual data test proved that the Motif-Hi-C method has a significant speed advantage over four mainstream Hi-C data analysis software. In addition, based on this innovative classification strategy, the present invention firstly performed upstream analysis on 7 groups of large-scale Hi-C data sets matched.fastq and unmatched.fastq file data, and explored the differences in the interaction information contained in the two parts of the data. In addition, the information of different levels of chromatin structure in these two parts of the data was deeply explored and compared through downstream analysis, revealing the influence of Motif distribution on Hi-C data. This innovative classification strategy also provides a new perspective for improving the quality of Hi-C data. Through this series of analyses, the efficiency and innovation of the Motif-Hi-C method in Hi-C data processing and analysis are demonstrated.

[0079] In summary, the present invention selected 5 small Hi-C datasets and 7 large Hi-C datasets for analysis. The above 12 datasets cover different cell types and restriction enzymes in Hi-C research, ensuring that the Motif-Hi-C method can be used to analyze different Hi-C datasets. The present invention compares the difference in running time between Motif-Hi-C and four Hi-C data analysis software in processing 12 Hi-C datasets. The running time results show that compared with the four Hi-C data analysis software, the Motif-Hi-C method can process Hi-C data more quickly. The ratio of the number of valid pairs to the total number of read pairs after deduplication is a key indicator for evaluating the quality of Hi-C data sets. By comparing the ratio of the number of valid interactions (valid interaction rmdup) to the total number of read pairs (total read pairs) obtained by Motif-Hi-C and four Hi-C data analysis software after processing 12 groups of Hi-C data sets, the results show that there is no significant difference between the indicators obtained using the Motif-Hi-C method and the other four Hi-C data analysis software, further indicating that it is feasible to use the Motif-Hi-C method to quickly judge the quality of Hi-C data. In addition, the present invention analyzes the changing trend of the Motif-Hi-C method and the four Hi-C data analysis software for 12 groups of data in the indicator of valid interaction rmdup / total read pairs. The results show that the changing trend of the Motif-Hi-C analysis results is very close to that of the other four Hi-C data analysis software. The above results show that the Motif-Hi-C method can provide a fast and accurate choice for quality control and usability analysis of Hi-C data, and when it is necessary to screen out the highest quality data set from multiple sets of Hi-C data. The Motif-Hi-C process of quickly judging the quality of Hi-C data can effectively achieve a preliminary and rapid assessment of the quality of the Hi-C data set, but because the alignment step is not performed on the reads data of the matched.fastq file, the exact location information of this part of the reads on the reference genome has not yet been determined, so it cannot be directly used for downstream chromatin structure analysis. After the Motif-Hi-C method is used to perform a preliminary assessment of the quality of Hi-C data, if the assessment results can be used for subsequent analysis, the matched.fastq file data is aligned, filtered, and deduplicated to obtain specific valid interaction reads information, and the matched.fastq and unmatched.fastq file data are compared in many aspects.At the same time, the present invention performs independent downstream analysis on the matched.fastq and unmatched.fastq data of the same Hi-C dataset, and further analyzes the potential association between Motif and different chromatin structures. The results show that compared with the reads data without Motif (unmatched.fastq), the reads data containing Motif (matched.fastq) show higher data quality and contain more chromatin structure information. The present invention not only retains the advantage of rapid assessment of dataset quality, but also expands the possibility of in-depth analysis of data, providing a flexible solution for efficient and accurate Hi-C data processing and analysis.

Claims

1. Motif-Hi-C, a Hi-C data quality assessment method based on proximity-linked Motif sequences, is characterized by: The method comprises the following steps: S1: Hi-C data storage fastq file for Motif matching: For single enzyme digestion Hi-C data, the KMP algorithm is used to match the Motif, and for multiple enzyme digestion Hi-C data, the AC automaton algorithm is used to match the Motif; S2: Classify the original Hi-C data storage fastq files: Classify the original Hi-C data storage fastq files into matched.fastq and unmatched.fastq files according to the restriction endonuclease cutting sites and Motif characteristics; S3: Noise reduction processing of unmatched.fastq file data: After performing alignment, filtering and deduplication steps on the unmatched.fastq file data, the noise-reduced unmatched.fastq file data and evaluation results are obtained; S4: simulated deduplication of matched.fastq file data: calculate the PCR amplification deduplication multiple according to the deduplication of unmatched.fastq file data, and then use the PCR amplification deduplication multiple calculated from the unmatched.fastq file data to simulate the PCR amplification deduplication of the matched.fastq file evaluation results; S5: Hi-C data quality assessment: The data assessment results of the unmatched.fastq file after noise reduction in step S3 are combined with the data assessment results of the matched.fastq file after simulated deduplication in step S4, and the ratio of the number of effective interactions to the total read pairs is calculated to evaluate the Hi-C data quality.

2. The method according to claim 1, characterized in that The method for classifying the original Hi-C data storage fastq files in step S2 comprises the following steps: a1: Screening read IDs containing specific motifs: Read the fastq files containing Hi-C sequencing reads. For each read in the R1.fastq and R2.fastq files, the program will search for read IDs containing specific motifs in their IDs. a2: Output qualified read IDs: For reads whose IDs contain specific motifs, the program will output their IDs into two new files respectively; a3: Merge and remove duplicate IDs: The program will merge the two folders in step a2 into a new file and rename it to ensure that all qualified read IDs are integrated. The merged ID list will then be deduplicated and saved as a new file. a4: Classify fastq files according to the ID list: Use the ID list of the file after deduplication in step a3 to classify the original R1.fastq and R2.fastq files. If the read ID exists in the file after deduplication in step a3, the read is classified as matched, and R1.matched.fastq and R2.matched.fastq files are generated for R1.fastq and R2.fastq, respectively; if the read ID does not exist in the file after deduplication in step a3, the read is classified as unmatched, and R1.unmatched.fastq and R2.unmatched.fastq files are generated for R1.fastq and R2.fastq, respectively.

3. The method according to claim 1, characterized in that The method for matching Motifs using the KMP algorithm for single enzyme digestion Hi-C data in step S1 comprises the following steps: 1) Build a partial match table based on the Motif sequence to quickly move the pointer during the matching process, and initialize a partial match table with the same length as the Motif, with all initial values ​​being 0; 2) Starting from the second character of the motif, calculate the longest common prefix and suffix length of each position in turn and record them in the partial match table. During the matching process, use the constructed partial match table to search for the motif in the read sequence; 3) Initialize two pointers to point to the start position of the reads sequence and Motif respectively; 4) Move the pointers in the reads sequence one by one, compare the base characters at the corresponding positions, and adjust the position of the motif according to the partial match table: if the current base character matches successfully, both pointers move back one position at the same time; If the current base character does not match and the value of the partial match table is greater than 0, the Motif is moved to the right and the pointer of the read sequence remains unchanged; if the current base character does not match and the value of the partial match table is 0, the pointers of the Motif and the read sequence are both moved one position to the right; 5) Repeat the above process until the match is completed or the pointer of the reads sequence moves to the end.

4. The method according to claim 1, characterized in that The method for matching Motifs using the AC automaton algorithm for multi-enzyme Hi-C data in step S1 comprises the following steps: 1) Construct a Trie tree as the search data structure of the AC automaton; 2) Construct a fail pointer so that when the current base character fails to match, it jumps to the base character with the longest common prefix and suffix to continue matching; 3) Scan the reads sequence for matching.

5. The method according to claim 1, characterized in that Bowtie2 comparison software is used for data comparison in step S3.

6. Application of the Hi-C data quality assessment method Motif-Hi-C based on proximity ligation Motif sequences as described in any one of claims 1 to 5 in Hi-C high-throughput sequencing data analysis.

7. A Hi-C high-throughput sequencing data analysis method, characterized in that: The method comprises the following steps: S1: Preliminary evaluation of Hi-C high-throughput sequencing data using the Hi-C data quality assessment method Motif-Hi-C based on proximity-linked Motif sequences described in any one of claims 1 to 5; S2: According to the results of the preliminary evaluation in step S1, if the evaluation results can be used for subsequent analysis, the matched.fastq file data is compared, filtered and deduplicated to obtain specific valid interaction reads information; S3: Perform downstream analysis on the matched.fastq file data processed in step S2 and the unmatched.fastq file data after noise reduction processing separately or combine them before performing downstream analysis.

8. Application of Motif-Hi-C, a Hi-C data quality assessment method based on proximity ligation Motif sequences, as described in any one of claims 1 to 5, in Hi-C data quality assessment.

9. A method for rapid analysis of large-scale Hi-C high-throughput sequencing data, characterized in that: The method comprises the following steps: S1: Hi-C data storage fastq file for Motif matching: For single enzyme digestion Hi-C data, the KMP algorithm is used to match the Motif, and for multiple enzyme digestion Hi-C data, the AC automaton algorithm is used to match the Motif; S2: Classify the original Hi-C data storage fastq files: Classify the original Hi-C data storage fastq files into matched.fastq and unmatched.fastq files according to the restriction endonuclease cutting sites and Motif characteristics; S3: Perform downstream analysis only on the matched.fastq data in step S2 after quality control.

10. The method according to claim 9, characterized in that The method for classifying the original Hi-C data storage fastq files in step S2 comprises the following steps: a1: Screening read IDs containing specific motifs: Read the fastq files containing Hi-C sequencing reads. For each read in the R1.fastq and R2.fastq files, the program will search for read IDs containing specific motifs in their IDs. a2: Output qualified read IDs: For reads whose IDs contain specific motifs, the program will output their IDs into two new files respectively; a3: Merge and remove duplicate IDs: The program will merge the two folders in step a2 into a new file and rename it to ensure that all qualified read IDs are integrated. The merged ID list will then be deduplicated and saved as a new file. a4: Classify fastq files according to the ID list: Use the ID list of the file after deduplication in step a3 to classify the original R1.fastq and R2.fastq files. If the read ID exists in the file after deduplication in step a3, the read is classified as matched, and R1.matched.fastq and R2.matched.fastq files are generated for R1.fastq and R2.fastq, respectively; if the read ID does not exist in the file after deduplication in step a3, the read is classified as unmatched, and R1.unmatched.fastq and R2.unmatched.fastq files are generated for R1.fastq and R2.fastq, respectively.

Images