Prediction method and prediction system for chromosome topological association domains

Through clustering algorithms, high-frequency interaction regions and kernels are identified and processed, and related kernel regions are merged, which solves the volatility and subjectivity problems of existing TAD prediction methods, and achieves a high accuracy and reliability chromosomal topological correlation domain. Predicted, suitable for different cell lines.

CN114446384BActive Publication Date: 2024-11-05CENT SOUTH UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210245600.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-14
Publication Date
2024-11-05
Estimated Expiration
2042-03-14

AI Technical Summary

Technical Problem

The existing chromosomal topological association domain (TAD) prediction methods are volatile and subjective, and cannot effectively utilize chromosome internal information, resulting in inaccurate identification results and the unique information needs of different cell lines increase the difficulty of research.

Method used

Use clustering algorithm to identify high-frequency interaction regions, judge and split the nucleus, process adjacent nucleus relationships, merge nucleus regions with high correlation, form predicted chromosomal topological correlation domains, reduce false positive results, and do not require pre-predictions. Define parameters.

Benefits of technology

High accuracy and reliability TAD prediction is achieved, false positive results are reduced, restrictive prediction of TAD attributes is avoided, and applicable to different cell lines.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114446384B_ABST
    Figure CN114446384B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for predicting a chromosome topological association domain, including obtaining each genomic block in an interaction matrix between genomic blocks and identifying a high-frequency interaction region; identifying a quasi-nucleus from the high-frequency interaction region for each genomic block: processing the quasi-nucleus identified on each chromosome to obtain non-overlapping quasi-nuclei; merging non-overlapping quasi-nuclei on a chromosome to obtain the nucleus of the chromosome topological association domain to be predicted; determining the subordinate relationship of each genomic block in the attachment candidate region and combining the nucleus of the chromosome topological association domain to obtain the final predicted chromosome topological association domain. The present invention also discloses a prediction system for realizing the method for predicting the chromosome topological association domain. The present invention makes full use of the global information of Hi-C data, reduces the scope of candidate boundary positioning, does not require the user to provide predefined parameters, can accurately predict the topological association domain, has high reliability, good accuracy and good effect.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of computer technology, and in particular relates to a prediction method and a prediction system for a chromosome topological association structural domain. Background Art

[0002] In recent years, the emergence of genome-wide chromosome configuration capture technology (High-throughput chromosome configuration capture technology, Hi-C) has promoted people's understanding of the spatial structural hierarchy of chromosomes. Relevant researchers converted the Hi-C sequencing data of mammalian cells into a Hi-C interaction matrix and visualized it, thereby discovering highly self-interactive regions at a resolution of less than 100kb. Such regions are topologically associated domains (TADs). The specific steps for constructing the Hi-C interaction matrix are: divide a chromosome into equal length N fragments and build into a N * N The matrix M , which is used to characterize the interaction signals between two fragments on a chromosome, where the unit-length fragments of equal length are called a genomic block, and the size of the genomic block is related to the resolution of the Hi-C interaction matrix. By counting the alignment of the sequencing fragment reads generated by the high-throughput chromosome conformation capture technology between genomic block pairs and the interaction frequency between N genomic blocks, the researchers constructed the Hi-C interaction matrix. For example, each sequencing fragment read can be split and aligned to a genomic block i and genomic blocks j , then in the matrix element M i,j , M j,i Add 1 to the total.

[0003] Currently, due to the limitations of microscopy and biotechnology, researchers are still unable to directly and completely observe TAD, and the formation mechanism of TAD is still a vague concept. Therefore, in order to obtain TAD information, some indirect methods must be used, such as using the interaction information between chromosome fragments captured by Hi-C sequencing data to construct a Hi-C interaction matrix, and then use related algorithms to predict TAD. In recent years, researchers have proposed methods based on machine learning algorithms to predict TAD; however, the application of these methods on different cell lines is greatly limited, because different cell lines often require a large amount of corresponding and unique related information to extract feature training models, which adds an additional burden to researchers.

[0004] Existing TAD prediction algorithms mainly predict TAD from the perspectives of interaction preference at the boundary, similarity within TAD, difference between TAD and non-TAD, change in contact frequency density within TAD, etc. These methods either focus only on finding the boundary and miss the information inside TAD; or need to use custom parameters to control the size of TAD, cluster termination threshold, local maximum, etc.; this makes the problem of identifying TAD very volatile and subjective; moreover, TAD, as an imprecisely defined structure, should not be predicted by restricting its own properties. Summary of the invention

[0005] One of the purposes of the present invention is to provide a method for predicting chromosome topological association domains with high reliability, good accuracy and good effect.

[0006] A second object of the present invention is to provide a prediction system for realizing the prediction method of the chromosome topological association domain.

[0007] The method for predicting the chromosome topological association domain provided by the present invention comprises the following steps:

[0008] S1. Obtain each genomic block in the interaction matrix between genomic blocks, and use clustering algorithm to identify the corresponding high-frequency interaction regions;

[0009] S2. For each genomic block, determine and identify from the corresponding high-frequency interaction regions whether there is a quasi-core centered on the genomic block:

[0010] If a quasi-core centered on the genomic block exists in the high-frequency interaction region, proceed to the subsequent steps;

[0011] If there is no quasi-core centered on the genomic block in the high-frequency interaction region, the high-frequency interaction region is split and then the quasi-core is re-judged and identified until the split region does not contain the genomic block;

[0012] S3. The quasi-nuclei identified on each chromosome are processed according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei;

[0013] S4. According to the correlation between the quasi-nuclei, the non-overlapping quasi-nuclei on a chromosome are merged, and the merged nuclei are used as the nuclei of the chromosome topological association domain to be predicted;

[0014] S5. Determine the affiliation of each genomic block in the appendix candidate region, and combine it with the core of the chromosome topological association domain obtained in step S4 to obtain the final predicted chromosome topological association domain.

[0015] The step S1 specifically uses the whole genome conformation capture technology and sequencing technology to obtain each genomic block in the interaction matrix between genomic blocks, and uses k =2 K The mean clustering algorithm is used for clustering to identify the corresponding high-frequency interaction areas.

[0016] The step S1 specifically includes the following steps:

[0017] S1.1. Use whole-genome conformation capture technology and sequencing technology to obtain the interaction matrix between genomic blocks;

[0018] S1.2. Assign 0 to the interaction value between each genomic block and itself on the diagonal of the interaction matrix between genomic blocks obtained in step S1.1;

[0019] S1.3. For any genomic block i ,use k =2 K Mean clustering algorithm for this genomic block i Clustering with other genomic blocks whose interaction values ​​are not 0;

[0020] S1.4. For each genomic block i Define the corresponding high-frequency interaction area ;in, l i Corresponding to genomic block i The minimum block number of genomic blocks in the high interaction class, r i Corresponding to genomic block i Maximum block number of genomic blocks in the high interaction class.

[0021] The following function is used as the classification function for other genomic blocks in step S1.3 :

[0022]

[0023] In the formula Genome Block i and genomic blocks j The interaction value of For the k The average of the centers; For taking and The function of operating on the category number corresponding to the nearest center; is the 2-norm; the initial center values ​​of the two classes and After setting the interaction value to non-zero in ascending order and The interaction value corresponding to the position, and Corresponding to the center of the low-frequency interaction class, The center corresponding to the high-frequency interaction class;

[0024] By solving the classification function, the class corresponding to the smallest distance from the center value is assigned to the genome block j .

[0025] The step S2 specifically includes the following steps:

[0026] S2.1. Computing genome blocks i High frequency interaction zone Submatrices composed in the interaction matrix between genomic blocks The average interaction value of

[0027] S2.2. Compare the average interaction value obtained in step S2.1 with the average interaction value of the adjacent 5 sub-matrices of the same window size:

[0028] If the average interaction value obtained in step S2.1 is greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction region is determined. To count genome blocks i of approval;

[0029] If the average interaction value obtained in step S2.1 is not greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction area is Split; re-judge and identify after splitting until the split area does not contain genomic blocks i Stop when

[0030] The adjacent 5 sub-matrices of the same window size are specifically the top 3 sub-matrices , and , the right submatrix , and a submatrix below .

[0031] High frequency interaction region Split; re-judge and identify after splitting until the split area does not contain genomic blocks i The specific steps include:

[0032] First, high-frequency interaction areas Interaction zone between medium and high frequency The genomic block with the smallest sum of interactions with other genomic blocks m i As the split point, the high-frequency interaction area High frequency interaction zone and high frequency interaction region ;

[0033] Then, make a judgment:

[0034] like i = m i , then it is determined that there is no genomic block i The quasi-nucleus of the center;

[0035] like i < m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1 to S2.2 to determine the quasi-nucleus;

[0036] like i > m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1~S2.2 to determine the quasi-nucleus.

[0037] The step S3 specifically includes the following steps:

[0038] S3.1. For each quasi-nucleus identified on a chromosome, determine the relationship between two adjacent quasi-nuclei:

[0039] If there is a containment relationship between two adjacent quasi-cores, the contained quasi-core is retained and the contained quasi-core is filtered;

[0040] If there is an overlapping relationship between two adjacent quasi-nuclei, a judgment is made again: if the two quasi-nuclei still meet the definition of quasi-nuclei after being merged, the two quasi-nuclei are merged into one quasi-nucleus; otherwise, the quasi-nucleus with the larger average interaction value among the two quasi-nuclei is retained, and the remaining quasi-nuclei are filtered;

[0041] S3.2. Repeat step S3.1 until all quasi-nuclei on the entire chromosome have been determined and processed, and finally non-overlapping quasi-nuclei are obtained.

[0042] The step S4 is specifically to calculate the cosine similarity between all adjacent quasi-nuclei, and merge several consecutive adjacent quasi-nuclei whose cosine similarity is higher than the set threshold and whose average interaction value between adjacent quasi-nuclei is greater than the average of the non-zero interaction values ​​on the entire chromosome into a new region, and use the region as the nucleus in the nucleus-accessory structure model of the chromosome topological association domain to be predicted.

[0043] The cosine similarity between all adjacent quasi-cores is calculated by using the following formula: pci and pc j Cosine similarity of :

[0044]

[0045] In the formula for pc i The eigenvector is the average interaction value of all other quasi-nuclei, and , , For approval pc k and pc i The average interaction value between for pc j The eigenvector is the average interaction value of all other quasi-kernels, and , , For approval pc k and pc j The average interaction value between is the inner product of vectors; The modulus of a vector.

[0046] The step S5 specifically defines the area between nuclei as the accessory region, determines the nucleus of the adjacent chromosome topological association structure domain to which each genomic block in each accessory region belongs, and thus obtains the final predicted chromosome topological association structure domain; each chromosome topological association structure domain includes a nucleus and the accessory regions on both sides of the nucleus.

[0047] The step S5 specifically includes the following steps:

[0048] S5.1. For two adjacent cores and The middle genomic block , filter out genomic blocks where the average interaction value of high-frequency interaction regions is less than the average of non-zero interaction values ​​on the entire chromosome;

[0049] S5.2. Based on step S5.1, and and the genomic block between the two nuclei The submatrix formed by the genomic blocks between two adjacent nuclei is formed by removing the background signal; the background signal is defined as the mean of the non-zero interaction values ​​in the submatrix formed by the genomic blocks between two adjacent nuclei;

[0050] S5.3. Based on step S5.2, and The middle genomic block , filter out genomic regions that are not present Any genomic block within has a non-zero interaction value;

[0051] S5.4. Based on step S5.3, calculate the two adjacent cores and Each remaining genomic block in the submatrix The average interaction value of the submatrix is ​​taken as the average interaction value of the submatrix, and the genomic block corresponding to the minimum average interaction value of the submatrix is ​​taken as the split point, and the genomic block upstream of the split point is identified as the core. The genomic block downstream of the split point is identified as the nucleus 's attachments; thus obtaining the final predicted chromosome topological association domain.

[0052] The present invention also provides a prediction system for realizing the prediction method of the chromosome topological association domain, comprising a high-frequency interaction region recognition module, a quasi-nucleus recognition module, a quasi-nucleus processing module, a chromosome topological association domain nucleus recognition module and a chromosome topological association domain recognition module which are serially connected in sequence; the high-frequency interaction region recognition module is used to obtain each genomic block in the interaction matrix between genomic blocks, and use a clustering algorithm to identify the corresponding high-frequency interaction region, and upload the obtained high-frequency interaction region to the quasi-nucleus recognition module; the quasi-nucleus recognition module is used to judge and identify whether there is a quasi-nucleus centered on the genomic block from the corresponding high-frequency interaction region for each genomic block, and upload the obtained quasi-nucleus to the quasi-nucleus processing module; the quasi-nucleus processing module is used to process each genomic block. The quasi-nuclei identified on the chromosome are processed according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei, and the obtained non-overlapping quasi-nuclei are uploaded to the chromosome topology association structure domain nucleus recognition module; the chromosome topology association structure domain nucleus recognition module is used to merge the non-overlapping quasi-nuclei on a chromosome according to the correlation between the quasi-nuclei, use the merged nucleus as the nucleus of the chromosome topology association structure domain to be predicted, and upload the obtained nucleus to the chromosome topology association structure domain recognition module; the chromosome topology association structure domain recognition module is used to determine the subordinate relationship of each genomic block in the accessory candidate area, and combine the received chromosome topology association structure domain nucleus to obtain the final predicted chromosome topology association structure domain, and output it.

[0053] The method and system for predicting chromosome topological association domains provided by the present invention make full use of the global information of Hi-C data, reduce the range of candidate boundary positioning, and thus reduce the occurrence of false positive results; at the same time, the present invention does not require the user to provide predefined parameters, so the present invention can accurately predict topological association domains with high reliability, good accuracy and good effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The figure is a schematic diagram of the method flow of the present invention.

[0055] Figure 2 The present invention is a schematic diagram of a process of an embodiment of the method.

[0056] Figure 3 It is a schematic diagram of the structure of the system of the present invention. DETAILED DESCRIPTION

[0057] like Figure 1 The method flow diagram of the method of the present invention is shown as follows: The method for predicting chromosome topological association domains provided by the present invention comprises the following steps:

[0058] S1. Obtain each genomic block in the interaction matrix between genomic blocks, and use clustering algorithm to identify the corresponding high-frequency interaction region; specifically, use whole genome conformation capture technology and sequencing technology to obtain each genomic block in the interaction matrix between genomic blocks (referred to as Hi-C interaction matrix), and use k =2 K The mean clustering algorithm is used for clustering to identify the corresponding high-frequency interaction areas;

[0059] The specific implementation includes the following steps:

[0060] S1.1. Use whole-genome conformation capture technology and sequencing technology to obtain the interaction matrix between genomic blocks;

[0061] S1.2. Assign 0 to the interaction value between each genomic block and itself on the diagonal of the interaction matrix between genomic blocks obtained in step S1.1;

[0062] S1.3. For any genomic block i ,use k =2 K Mean clustering algorithm for this genomic block i Clustering is performed on other genomic blocks whose interaction values ​​are not 0; the following function is used as the classification function for other genomic blocks :

[0063]

[0064] In the formula Genome Block i and genomic blocks j The interaction value of For the k The average of the centers; For taking and The function of operating on the category number corresponding to the nearest center; is the 2-norm; the initial center values ​​of the two classes and After setting the interaction value to non-zero in ascending order and The interaction value corresponding to the position, and Corresponding to the center of the low-frequency interaction class, The center corresponding to the high-frequency interaction class;

[0065] By solving the classification function, the class corresponding to the smallest distance from the center value is assigned to the genome block j ;

[0066] S1.4. For each genomic block i Define the corresponding high-frequency interaction area ;in, l i Corresponding to genomic block i The minimum block number of genomic blocks in the high interaction class, r i Corresponding to genomic block i The maximum block number of the genomic blocks in the high interaction class;

[0067] S2. For each genomic block, determine and identify from the corresponding high-frequency interaction regions whether there is a quasi-core centered on the genomic block:

[0068] If a quasi-core centered on the genomic block exists in the high-frequency interaction region, proceed to the subsequent steps;

[0069] If there is no quasi-core centered on the genomic block in the high-frequency interaction region, the high-frequency interaction region is split and then the quasi-core is re-judged and identified until the split region does not contain the genomic block;

[0070] The specific implementation includes the following steps:

[0071] S2.1. Computing genome blocks i High frequency interaction zone Submatrices composed in the interaction matrix between genomic blocks The average interaction value of

[0072] S2.2. Compare the average interaction value obtained in step S2.1 with the average interaction value of the adjacent 5 sub-matrices of the same window size:

[0073] If the average interaction value obtained in step S2.1 is greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction region is determined. To count genome blocks i of approval;

[0074] If the average interaction value obtained in step S2.1 is not greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction area is Split; re-judge and identify after splitting until the split area does not contain genomic blocks i Stop when

[0075] The adjacent 5 sub-matrices of the same window size are specifically the top 3 sub-matrices , and , the right submatrix , and a submatrix below ;

[0076] High frequency interaction region Split; re-judge and identify after splitting until the split area does not contain genomic blocks i The specific steps include:

[0077] First, high-frequency interaction areas Interaction zone between medium and high frequency The genomic block with the smallest sum of interactions with other genomic blocks m i As the split point, the high-frequency interaction area High frequency interaction zone and high frequency interaction region ;

[0078] Then, make a judgment:

[0079] like i = m i , then it is determined that there is no genomic block i The quasi-nucleus of the center;

[0080] like i < m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1 to S2.2 to determine the quasi-nucleus;

[0081] like i > m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1 to S2.2 to determine the quasi-nucleus;

[0082] S3. The quasi-nuclei identified on each chromosome are processed according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei; specifically, the steps include:

[0083] S3.1. For each quasi-nucleus identified on a chromosome, determine the relationship between two adjacent quasi-nuclei:

[0084] If there is a containment relationship between two adjacent quasi-cores, the contained quasi-core is retained and the contained quasi-core is filtered;

[0085] If there is an overlapping relationship between two adjacent quasi-nuclei, a judgment is made again: if the two quasi-nuclei still meet the definition of quasi-nuclei after being merged, the two quasi-nuclei are merged into one quasi-nucleus; otherwise, the quasi-nucleus with the larger average interaction value among the two quasi-nuclei is retained, and the remaining quasi-nuclei are filtered;

[0086] S3.2. Repeat step S3.1 until all quasi-nuclei on the entire chromosome have been determined and processed, and finally non-overlapping quasi-nuclei are obtained;

[0087] S4. According to the correlation between the quasi-nuclei, the non-overlapping quasi-nuclei on a chromosome are merged, and the merged nuclei are used as the nuclei of the chromosome topological association domain (TAD) to be predicted; specifically, the cosine similarity between all adjacent quasi-nuclei is calculated, and a number of consecutive adjacent quasi-nuclei whose cosine similarity is higher than the set threshold and whose average interaction value between adjacent quasi-nuclei is higher than the average of the non-zero interaction values ​​on the entire chromosome are merged into a new region, and the region is used as the nucleus in the nucleus-accessory structure model of the chromosome topological association domain to be predicted;

[0088] In specific implementation, the following formula is used to calculate the adjacent quasi-core pc i and pc j Cosine similarity of :

[0089]

[0090] In the formula for pc i The eigenvector is the average interaction value of all other quasi-kernels, and , , For approval pc k and pc i The average interaction value between for pc j The eigenvector is the average interaction value of all other quasi-kernels, and , , For approval pc k and pc j The average interaction value between is the inner product of vectors; is the modulus of a vector;

[0091] S5. Determine the subordinate relationship of each genomic block in the accessory candidate region, and combine the core of the chromosome topological association structure domain obtained in step S4 to obtain the final predicted chromosome topological association structure domain; specifically, define the area between the cores as the accessory region, determine the core of the adjacent chromosome topological association structure domain to which each genomic block in each accessory region belongs, so as to obtain the final predicted chromosome topological association structure domain; each chromosome topological association structure domain includes a core and the accessory regions on both sides of the core;

[0092] The specific implementation includes the following steps:

[0093] S5.1. For two adjacent cores and The middle genomic block , filter out genomic blocks where the average interaction value of high-frequency interaction regions is less than the average of non-zero interaction values ​​on the entire chromosome;

[0094] S5.2. Based on step S5.1, and and the genomic block between the two nuclei The submatrix formed by the genomic blocks between two adjacent nuclei is formed by removing the background signal; the background signal is defined as the mean of the non-zero interaction values ​​in the submatrix formed by the genomic blocks between two adjacent nuclei;

[0095] S5.3. Based on step S5.2, and The middle genomic block , filter out genomic regions that are not present Any genomic block within has a non-zero interaction value;

[0096] S5.4. Based on step S5.3, calculate the two adjacent cores and Each remaining genomic block in the submatrix The average interaction value of the submatrix is ​​taken as the average interaction value of the submatrix, and the genomic block corresponding to the minimum average interaction value of the submatrix is ​​taken as the split point, and the genomic block upstream of the split point is identified as the core. The genomic block downstream of the split point is identified as the nucleus 's attachments; thus obtaining the final predicted chromosome topological association domain.

[0097] The method of the present invention is further described below in conjunction with an embodiment:

[0098] like Figure 2 The method for predicting chromosome topological association domains based on the nuclear-accessory structure model provided in the embodiment includes the following steps; the Hi-C map in the figure is a Hi-C interaction matrix of GM12878_combined after KR normalization at a resolution of 50 kb contained in the GSE63525 data set, and the specific segment is the 120-200 genomic block of chromosome 1;

[0099] Step S1, for each genomic block in the interaction matrix between genomic blocks obtained by whole genome conformation capture technology and sequencing technology (referred to as Hi-C interaction matrix), a K-means clustering method is used to identify its high-frequency interaction region;

[0100] like Figure 2 -①As shown ( Figure 2 -① is the preprocessing process of Hi-C interaction matrix), the interaction value between each genomic block and itself on the diagonal of the Hi-C interaction matrix of GM12878_combined after KR normalization at 50kb resolution is assigned to 0;

[0101] like Figure 2 -②As shown ( Figure 2 -② is the identification process of high-frequency interaction regions), for each genomic block i ,use k =2 K-means clustering algorithm is used to perform clustering on other genomic blocks with non-zero interaction values. k =2, the classification functions of other genomic blocks are:

[0102]

[0103] in, Genome Block i and j The interaction value of It is k The average of the centers. The initial center values ​​of the two classes and Set to non-zero interaction value after sorting in ascending order and The interaction value corresponding to the position, Corresponding to the center of the low-frequency interaction class, The center of the corresponding high-frequency interaction class; by solving the classification function, the class corresponding to the smallest distance from the center value is assigned to the genome block j ;

[0104] For each genomic block i Define its high frequency interaction region ( l i , r i ), l i Corresponding genomic block i The minimum block number of genomic blocks in the high interaction class, r i Corresponding genomic block i The maximum block number of the genomic block in the high interaction class; the schematic diagram of the high frequency interaction region is as follows Figure 2 -②b shown;

[0105] Step S2: Figure 2 -③a as shown ( Figure 2 -③ is the construction process of TADs quasi-core). For each genomic block, determine and identify whether there is a quasi-core centered on the genomic block from its high-frequency interaction region;

[0106] Quasi-nuclei are defined as those genomic blocks i High frequency interaction zone Submatrices composed in the Hi-C interaction matrix The average interaction value of is greater than that of the 5 adjacent sub-matrices of the same window size, including the 3 sub-matrices above , and , a submatrix on the right , and the following submatrix , then the high frequency interaction region The genome block i of approval;

[0107] If the genomic block i High frequency interaction region Submatrices composed in the Hi-C interaction matrix If the average interaction value of is not greater than that of the other five adjacent sub-matrices of the same window size, then the high-frequency interaction area After the split, the quasi-nuclei are re-judged and identified until the split region does not contain genomic blocks. i Only then stopped;

[0108] When splitting: for genomic blocks i High frequency interaction region To split, first use the high-frequency interaction area The genomic block with the smallest sum of interactions with other genomic blocks in the medium-high frequency interaction region m i As the split point, the high-frequency interaction area Divided into two high-frequency interaction zones and ;

[0109] Furthermore, when i = m i , then it is determined that there is no genomic block i The quasi-nucleus of the center; i < m i , then continue to Re-judge and identify the approval; when i > m i , then continue to Re-judge and identify the approval; the judgment and identification process is as shown above;

[0110] Step S3: Figure 2 -③ As shown in b and c, the quasi-nuclei identified on each chromosome are filtered or merged according to the inclusion or overlap relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei;

[0111] When there is a containment relationship between two adjacent quasi-cores, the contained quasi-core is retained and the contained quasi-core is filtered;

[0112] When two adjacent quasi-nuclei overlap, if the two can still meet the definition of quasi-nuclei when combined, they will be combined into one quasi-nucleus; otherwise, only the quasi-nucleus with a larger average interaction frequency will be retained.

[0113] After processing a group of two-by-two adjacent quasi-nuclei, start from the next quasi-nucleus to search for two-by-two adjacent, contained or overlapping quasi-nuclei and perform the same processing until there are quasi-nuclei that do not overlap each other on the entire chromosome;

[0114] Step S4: Figure 2 -④As shown ( Figure 2 -④ is the process of constructing the nucleus in the nucleus-accessory structure model of TADs). According to the correlation between quasi-nuclei, the non-overlapping quasi-nuclei on a chromosome are merged, and the merged nucleus is regarded as the nucleus of the chromosome topologically associated domain (TADs) to be predicted;

[0115] Use cosine similarity to identify all adjacent pairwise quasi-kernels pc i and pc j The correlation calculation formula is as follows: ;

[0116] Set a correlation threshold, and merge two or more consecutive adjacent quasi-nuclei with a similarity higher than the threshold and an average interaction value between adjacent quasi-nuclei greater than the average of the non-zero interaction values ​​on the entire chromosome into a new region as the nucleus in the nucleus-attachment structure model of a TAD.

[0117] Step S5: Figure 2 -⑤As shown ( Figure 2 -⑤ is the process of establishing a complete core-attachment structure model of TADs), the area between the cores is defined as the attachment candidate area, and each genomic block in the attachment candidate area is determined to which adjacent core each genomic block belongs. Finally, each TAD predicted consists of a core and its attachments on both sides; the specific implementation includes the following steps:

[0118] S5.1. For two adjacent cores and The middle genomic block , filter out genomic blocks where the average interaction value of high-frequency interaction regions is less than the average of non-zero interaction values ​​on the entire chromosome;

[0119] S5.2. Based on step S5.1, and and the genomic block between the two nuclei The submatrix formed by the genomic blocks between two adjacent nuclei is formed by removing the background signal; the background signal is defined as the mean of the non-zero interaction values ​​in the submatrix formed by the genomic blocks between two adjacent nuclei;

[0120] S5.3. Based on step S5.2, and The middle genomic block , filter out genomic regions that are not present Any genomic block within has a non-zero interaction value;

[0121] S5.4. Based on step S5.3, calculate the two adjacent cores and Each remaining genomic block in the submatrix The average interaction value of the submatrix is ​​taken as the average interaction value of the submatrix, and the genomic block corresponding to the minimum average interaction value of the submatrix is ​​taken as the split point, and the genomic block upstream of the split point is identified as the core. The genomic block downstream of the split point is identified as the nucleus 's attachments; thus obtaining the final predicted chromosome topological association domain.

[0122] like Figure 3 The structure schematic diagram of the system of the present invention is shown as follows: The present invention also provides a prediction system for realizing the prediction method of the chromosome topological association domain, comprising a high-frequency interaction region recognition module, a quasi-nucleus recognition module, a quasi-nucleus processing module, a chromosome topological association domain nucleus recognition module and a chromosome topological association domain recognition module which are serially connected in sequence; the high-frequency interaction region recognition module is used to obtain each genomic block in the interaction matrix between genomic blocks, and use a clustering algorithm to identify the corresponding high-frequency interaction region, and upload the obtained high-frequency interaction region to the quasi-nucleus recognition module; the quasi-nucleus recognition module is used to judge and identify whether there is a quasi-nucleus centered on the genomic block from the corresponding high-frequency interaction region for each genomic block, and upload the obtained quasi-nucleus to the quasi-nucleus processing module; the quasi-nucleus processing module is used to judge and identify whether there is a quasi-nucleus centered on the genomic block from the corresponding high-frequency interaction region for each genomic block, and upload the obtained quasi-nucleus to the quasi-nucleus processing module; The processing module is used to process the quasi-nuclei identified on each chromosome according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei, and upload the obtained non-overlapping quasi-nuclei to the chromosome topology association structure domain nucleus recognition module; the chromosome topology association structure domain nucleus recognition module is used to merge the non-overlapping quasi-nuclei on a chromosome according to the correlation between the quasi-nuclei, use the merged nucleus as the nucleus of the chromosome topology association structure domain to be predicted, and upload the obtained nucleus to the chromosome topology association structure domain recognition module; the chromosome topology association structure domain recognition module is used to determine the subordinate relationship of each genomic block in the accessory candidate area, and combine the received chromosome topology association structure domain nucleus to obtain the final predicted chromosome topology association structure domain, and output it.

Claims

1. A method for predicting chromosome topological association domains, characterized in that The steps include: S1. Obtain each genomic block in the interaction matrix between genomic blocks, and use clustering algorithm to identify the corresponding high-frequency interaction regions; S2. For each genomic block, determine and identify from the corresponding high-frequency interaction regions whether there is a quasi-core centered on the genomic block: If there is a quasi-core centered on the genomic block in the high-frequency interaction region, proceed to the subsequent steps; If there is no quasi-core centered on the genomic block in the high-frequency interaction region, the high-frequency interaction region is split and then the quasi-core is re-judged and identified until the split region does not contain the genomic block; S3. The quasi-nuclei identified on each chromosome are processed according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei; S4. According to the correlation between the quasi-nuclei, the non-overlapping quasi-nuclei on a chromosome are merged, and the merged nuclei are used as the nuclei of the chromosome topological association domain to be predicted; S5. Determine the affiliation of each genomic block in the appendix candidate region, and combine it with the core of the chromosome topological association domain obtained in step S4 to obtain the final predicted chromosome topological association domain.

2. The method for predicting chromosome topological association domains according to claim 1, characterized in that The step S1 specifically uses the whole genome conformation capture technology and sequencing technology to obtain each genomic block in the interaction matrix between genomic blocks, and uses k =2 K The mean clustering algorithm is used for clustering to identify the corresponding high-frequency interaction areas.

3. The method for predicting chromosome topological association domains according to claim 2, characterized in that The step S1 specifically includes the following steps: S1.

1. Use whole-genome conformation capture technology and sequencing technology to obtain the interaction matrix between genomic blocks; S1.

2. Assign 0 to the interaction value between each genomic block and itself on the diagonal of the interaction matrix between genomic blocks obtained in step S1.1; S1.

3. For any genomic block i ,use k =2 K Mean clustering algorithm for this genomic block i Cluster the other genomic blocks whose interaction values ​​are not 0; use the following function as the classification function of other genomic blocks in step S1.3 : In the formula Genome Block i and genomic blocks j The interaction value of For the k The average of the centers; For taking and The function of operating on the category number corresponding to the nearest center; is the 2-norm; the initial center values ​​of the two classes and After setting the interaction value to non-zero in ascending order and The interaction value corresponding to the position, and Corresponding to the center of the low-frequency interaction class, The center corresponding to the high-frequency interaction class; By solving the classification function, the class corresponding to the smallest distance from the center value is assigned to the genome block. j ; S1.

4. For each genomic block i Define the corresponding high-frequency interaction area ; in, l i Corresponding to genomic block i The minimum block number of genomic blocks in the high interaction class, r i Corresponding to genomic block i Maximum block number of genomic blocks in the high interaction class.

4. The method for predicting chromosome topological association domains according to claim 3, characterized in that The step S2 specifically includes the following steps: S2.

1. Computing genome blocks i High frequency interaction zone Submatrices composed in the interaction matrix between genomic blocks The average interaction value of S2.

2. Compare the average interaction value obtained in step S2.1 with the average interaction value of the adjacent 5 sub-matrices of the same window size: If the average interaction value obtained in step S2.1 is greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction region is determined. To count genome blocks i of approval; If the average interaction value obtained in step S2.1 is not greater than the average interaction value of the adjacent 5 sub-matrices with the same window size, the high-frequency interaction area is To split; After splitting, re-judgment and identification are performed until the split area does not contain genomic blocks i Stop when The adjacent 5 sub-matrices of the same window size are specifically the top 3 sub-matrices , and , the right submatrix , and a submatrix below .

5. The method for predicting chromosome topological association domains according to claim 4, characterized in that High frequency interaction region Split; re-judge and identify after splitting until the split area does not contain genomic blocks i The specific steps include: First, high-frequency interaction areas Interaction zone between medium and high frequency The genomic block with the smallest sum of interactions with other genomic blocks m i As the split point, the high-frequency interaction area High frequency interaction zone and high frequency interaction region ; Then, make a judgment: like i = m i , then it is determined that there is no genomic block i The quasi-nucleus of the center; like i < m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1 to S2.2 to determine the quasi-nucleus; like i > m i , then the high frequency interaction area As the high-frequency interaction region where the genomic block is located, repeat steps S2.1~S2.2 to determine the quasi-nucleus.

6. The method for predicting chromosome topological association domains according to claim 5, characterized in that The step S3 specifically includes the following steps: S3.

1. For each quasi-nucleus identified on a chromosome, determine the relationship between two adjacent quasi-nuclei: If there is a containment relationship between two adjacent quasi-cores, the contained quasi-core is retained and the contained quasi-core is filtered; If there is an overlapping relationship between two adjacent quasi-nuclei, a judgment is made again: if the two quasi-nuclei still meet the definition of quasi-nuclei after being merged, the two quasi-nuclei are merged into one quasi-nucleus; otherwise, the quasi-nucleus with the larger average interaction value among the two quasi-nuclei is retained, and the remaining quasi-nuclei are filtered; S3.

2. Repeat step S3.1 until all quasi-nuclei on the entire chromosome have been determined and processed, and finally non-overlapping quasi-nuclei are obtained.

7. The method for predicting chromosome topological association domains according to claim 6, characterized in that The step S4 is specifically to calculate the cosine similarity between all adjacent quasi-nuclei, and merge several consecutive adjacent quasi-nuclei whose cosine similarity is higher than the set threshold and whose average interaction value between adjacent quasi-nuclei is greater than the average of the non-zero interaction values ​​on the entire chromosome into a new region, and use the region as the nucleus in the nucleus-accessory structure model of the chromosome topological association domain to be predicted.

8. The method for predicting chromosome topological association domains according to claim 7, characterized in that The step S5 specifically defines the area between nuclei as the accessory region, determines the nucleus of the adjacent chromosome topological association structure domain to which each genomic block in each accessory region belongs, and thus obtains the final predicted chromosome topological association structure domain; each chromosome topological association structure domain includes a nucleus and the accessory regions on both sides of the nucleus.

9. The method for predicting chromosome topological association domains according to claim 8, characterized in that The step S5 specifically includes the following steps: S5.

1. For two adjacent cores and The middle genomic block , filter out genomic blocks where the average interaction value of high-frequency interaction regions is less than the average of non-zero interaction values ​​on the entire chromosome; S5.

2. Based on step S5.1, and and the genomic block between the two nuclei The submatrix formed by the genomic blocks between two adjacent nuclei is formed by removing the background signal; the background signal is defined as the mean of the non-zero interaction values ​​in the submatrix formed by the genomic blocks between two adjacent nuclei; S5.

3. Based on step S5.2, and The middle genomic block , filter out genomic regions that are not present Any genomic block within has a non-zero interaction value; S5.

4. Based on step S5.3, calculate the two adjacent cores and Each remaining genomic block in the submatrix The average interaction value of the submatrix is ​​taken as the average interaction value of the submatrix, and the genomic block corresponding to the minimum average interaction value of the submatrix is ​​taken as the split point, and the genomic block upstream of the split point is identified as the core. The genomic block downstream of the split point is identified as the nucleus 's attachments; thus obtaining the final predicted chromosome topological association domain.

10. A prediction system for implementing the method for predicting chromosome topological association domains according to any one of claims 1 to 9, characterized in that It includes a high-frequency interaction region recognition module, a quasi-nuclear recognition module, a quasi-nuclear processing module, a chromosome topology association domain nucleus recognition module and a chromosome topology association domain recognition module which are connected in series in sequence; the high-frequency interaction region recognition module is used to obtain each genomic block in the interaction matrix between genomic blocks, use a clustering algorithm to identify the corresponding high-frequency interaction region, and upload the obtained high-frequency interaction region to the quasi-nuclear recognition module; The quasi-nucleus identification module is used to determine and identify, for each genomic block, from the corresponding high-frequency interaction region whether there is a quasi-nucleus centered on the genomic block, and upload the obtained quasi-nucleus to the quasi-nucleus processing module; The quasi-nucleus processing module is used to process the quasi-nuclei identified on each chromosome according to the relationship between two adjacent quasi-nuclei to obtain non-overlapping quasi-nuclei, and upload the obtained non-overlapping quasi-nuclei to the chromosome topology association structure domain nucleus recognition module; the chromosome topology association structure domain nucleus recognition module is used to merge the non-overlapping quasi-nuclei on a chromosome according to the correlation between the quasi-nuclei, use the merged nucleus as the nucleus of the chromosome topology association structure domain to be predicted, and upload the obtained nucleus to the chromosome topology association structure domain recognition module; the chromosome topology association structure domain recognition module is used to determine the subordinate relationship of each genomic block in the accessory candidate area, and combine the received chromosome topology association structure domain nucleus to obtain the final predicted chromosome topology association structure domain, and output it.