Methods for detecting structural variation in genomic regions
By extracting and processing SR and PE signals in high-throughput sequencing data, combining genomic data and biological characteristics, detecting structural variations in genomic regions, solving the false positive problem in the prior art and improving the sensitivity and accuracy of detection , suitable for precision medicine.
Patent Information
- Application Number
- CN202011162206.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-10-27
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2040-10-27
AI Technical Summary
The existing high-throughput sequencing technology has false positive problems in detecting genomic structural mutations, and lacks high-sensitivity and high-precision detection methods, which affects its application in the field of precision medicine.
Signal clustering and filtering is performed by extracting and processing Split-reads (SR) and Paired-end (PE) signals in high-throughput sequencing data, combining genomic data and biological characteristics to detect structural variations in genomic regions, including For missing, repeat, inverted and translocation, dynamically set the threshold for insertion fragment length outliers to adapt to different data sources.
Improve the sensitivity and accuracy of structural variation detection, can more accurately detect structural variation in genomic regions, is suitable for the field of precision medicine, and supports detection of structural variation in genes, adapting to different sample types, including poor quality FFPE tissues sample.
Smart Images

Figure CN112349346B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of bioinformatics, and in particular relates to a method and system for detecting structural variations (SVs) in genomic regions using high-throughput sequencing data. Technical Background
[0002] Structural variants (SVs) are one of the main forms of variation in the human genome. Structural variations include deletion, duplication, inversion, and translocation of chromosome segments. Structural variations exist in both normal and tumor genomes, although they occur more frequently in tumor genomes. Structural variations in some genes may be associated with genetic risk and sensitivity to targeted therapy.
[0003] High-throughput sequencing technology, also known as the next-generation sequencing technology (NGS), provides convenience for low-cost, large-scale detection of structural variations. In addition, the detection method based on the second-generation sequencing technology can detect structural variations with breakpoint information at the base resolution level, while other low-throughput detection technologies such as fluorescence in situ hybridization (FISH) can only detect whether a certain structural variation exists, and cannot obtain breakpoint information. Microarray comparative genomic hybridization (array CGH) using microarray technology can only detect unbalanced structural variations including deletions or duplications, which is often used to detect copy number variations (CNV). In addition, the resolution of the breakpoint information it provides is also very limited.
[0004] As a complex form of variation, structural variation requires effective integration of multiple signals in NGS data to be more accurately detected. Currently, the detection of structural variation mainly relies on three signals: paired-end reads signal (PE for short), split-reads signal (SR for short) and read depth signal (RD for short). At present, structural variation detection software that uses one or more of the above signals has significant false positive problems, which greatly affects the application of structural variation detection in the field of precision medicine.
[0005] In summary, the field currently lacks highly sensitive and accurate structural variation detection methods based on high-throughput sequencing technology for application in the field of precision medicine. Summary of the invention
[0006] Listed below are some of the terms and their definitions used in this disclosure.
[0007] NGS: High-Throughput Sequencing, also known as Next Generation Sequencing (NGS), is relative to traditional Sanger Sequencing.
[0008] FASTQ format file: a text format that saves biological sequences (usually nucleic acid sequences) and their sequencing quality score information.
[0009] SAM format file: SAM (Sequence Alignment / Map format) file is a sequence alignment format standard developed by Sanger. It is a text format with TAB as the delimiter.
[0010] BAM format file: A binary file of SAM format file.
[0011] BED format file: BED (Browser Extensible Data) file is a format that defines track feature information such as comments. It is a text format with TAB as the delimiter.
[0012] SV: structural variation.
[0013] Insert size: The length of the sequence between the two end adapters. When a pair of reads originates from a structural variation event, they often have abnormal insert lengths after being posted back to the genome by alignment software.
[0014] Soft-clipping: refers to a part of an alignment record (read length) that cannot be posted back to the genome, but it is still part of the alignment record.
[0015] Split-read: refers to a comparison record (read length) with multiple replies, and different parts of the comparison record can be posted to different places in the genome. Read lengths with multiple replies may indicate the presence of structural variation.
[0016] Paired-end: refers to the two reads of double-end sequencing, that is, a pair of reads. Two reads with abnormal directions or abnormal insert lengths may indicate the presence of structural variations.
[0017] QNAME: Read name. Two reads of paired-end sequencing have the same QNAME.
[0018] SAM / BAM FLAG: The FLAG field of the SAM / BAM file is used to record information such as whether to reply, reply direction, and double-end pairing.
[0019] The present disclosure provides a method for detecting structural variation in a genomic region based on high-throughput sequencing technology, which can detect structural variation in genes, between genes and within genes, including deletions, duplications, inversions and translocations, etc. The method mainly includes the steps of signal extraction, signal processing and report generation. In the signal extraction stage, a sequence alignment file of high-throughput sequencing data, such as a SAM file or a BAM file, is read to extract SR signals and / or PE signals.
[0020] Then, in the signal processing stage, the SR signal and / or PE signal are used to perform steps such as determining the type of structural variation, signal clustering, and signal filtering. After the processing is completed, the two signal clusters are optionally merged based on the structural variation to form a comprehensive signal cluster, and filtered using a filtering system based on sequencing data, genomic data, and biological characteristics. Finally, the abundance of structural variation is optionally estimated and a report of the corresponding format file is output.
[0021] Accordingly, in one aspect, the present disclosure relates to a method of detecting a structural variation (SV) in a genomic region, the method comprising:
[0022] a. Read the sequence alignment file of high-throughput sequencing data,
[0023] i) The read length containing soft-clipping is used as the SR (Split-read) signal to enter the SR signal analysis process; and / or
[0024] ii) read pairs whose re-attachment direction in the reference genome is not forward-reverse, and read pairs whose re-attachment direction in the reference genome is forward-reverse but with abnormal insert length (insert size) are entered into the PE signal analysis process as PE (Paired-end) signals;
[0025] b. Performing an SR signal analysis process to mark the SV type, and clustering the SR signals marked as the same SV event to form an SR signal cluster; and / or performing a PE signal analysis process to mark the SV type, and clustering the PE signals marked as the same SV event to form a PE signal cluster;
[0026] c. filtering the SR signal cluster and / or the PE signal cluster to obtain a filtered signal cluster indicating an SV in a genomic region; and optionally
[0027] d. Output result report.
[0028] In another aspect, the present disclosure relates to a method of detecting a structural variation (SV) in a genomic region, the method comprising:
[0029] a. Read the sequence alignment file of high-throughput sequencing data, and enter the SR signal analysis process with the read length containing soft-clipping as the SR signal; enter the PE signal analysis process with the read length pairs whose re-attachment direction in the reference genome is not forward-reverse, and the read length pairs whose re-attachment direction in the reference genome is forward-reverse but with abnormal insertion fragment length as PE signals;
[0030] b. Performing the SR signal analysis process and the PE signal analysis process respectively to mark the SV type, and clustering the SR signals marked as the same SV event to form an SR signal cluster, and clustering the PE signals marked as the same SV event to form a PE signal cluster;
[0031] c. merging the SR signal cluster and the PE signal cluster marked as the same SV event according to the breakpoint coordinate information and the QNAME of the read length to form a comprehensive signal cluster, wherein the comprehensive signal cluster includes the merged SR signal cluster and the PE signal cluster, or the separate SR signal cluster or the PE signal cluster that failed to be merged;
[0032] d. filtering the integrated signal clusters to obtain filtered integrated signal clusters indicating SVs in the genomic region; and optionally
[0033] e. Output result report.
[0034] Figure 1 A flow diagram of an exemplary method of the present disclosure is shown.
[0035] In some embodiments, the sequence alignment file can be selected from a SAM file and a BAM file.
[0036] Due to differences in library construction methods, sequencing sample types, etc., the insert length value distribution of high-throughput sequencing data from different sources may be quite different. Therefore, in some cases, a single preset insert length outlier threshold may not be suitable for data with different methods or sources. In some embodiments of the method disclosed herein, the insert length outlier threshold can be dynamically set by calculating the insert length distribution.
[0037] In some embodiments, the read length having an abnormal insert length value may be determined by the following steps:
[0038] Read the insert length values of all reads in the sequence alignment file, or randomly read the insert length values of a certain number of reads;
[0039] constructing an insert length distribution using the insert length values, and determining an outlier threshold value significantly above the insert length distribution; and
[0040] Reads having insert lengths exceeding the outlier threshold are determined.
[0041] In some embodiments, the read length having an abnormal insert length value may be determined by the following steps:
[0042] Read the insert length values of all reads in the sequence alignment file, or randomly read the insert length values of a certain number of reads;
[0043] constructing a normal or log-normal distribution using the insert length values, and calculating parameters μ and σ of the normal or log-normal distribution; and
[0044] The insert fragment length value exceeding μ+Nσ is determined as having an abnormal insert fragment length value,
[0045] Wherein N is an integer between 3 and 10. For example, N can be 3, 4, 5, 6, 7, 8, 9 or 10.
[0046] In some embodiments, the insert length values of all reads in the sequence alignment file are read to construct a normal distribution or a log-normal distribution. In other embodiments, when the number of reads in the sequence alignment file is large, a certain number of insert length values of the reads can be randomly read for calculation to reduce the calculation time of the insert length distribution parameters. There is no limit to the number of randomly read reads, as long as it is sufficient to construct a normal distribution or a log-normal distribution and calculate its distribution parameters.
[0047] In some exemplary embodiments, the number of read lengths for random reads can be hundreds to millions, for example, at least 1,000, at least 5,000, at least 10,000, at least 50,000, at least 100,000, at least 500,000 or at least 1 million reads, or can be no more than 5 million, no more than 2 million, no more than 1 million or no more than 500,000 reads.
[0048] In some embodiments, a normal distribution is used to calculate the threshold for insert length outliers. In other embodiments, a log-normal distribution is used to calculate the threshold for insert length outliers.
[0049] In some embodiments, the SR signal analysis process may include:
[0050] The reads entering the SR signal analysis process were grouped according to QNAME (i.e., reads with the same QNAME were grouped into the same group), and in each group, the primary and secondary re-postions were paired according to the FLAG values of the reads, and the reads that could not be paired were filtered out;
[0051] Mark SV type, including:
[0052] In cases where the paired primary and secondary reattachments were located on different chromosomes, the SV type was marked as a translocation;
[0053] In the case where the paired primary reattachment and the suboptimal reattachment are located on the same chromosome, the primary reattachment and the suboptimal reattachment are sorted according to the chromosome coordinates to obtain the sorted reattachment direction information, which has four possibilities: forward-reverse, reverse-forward, forward-forward and reverse-reverse; at the same time, the soft-clipping position information of the primary reattachment and the suboptimal reattachment after sorting is obtained, which has four possibilities: left-right, right-left, left-left and right-right;
[0054] If the reclipping direction is forward-forward or reverse-reverse, and the soft-clipping position is right-left, the SV type is marked as missing;
[0055] If the reclipping direction is forward-forward or reverse-reverse, and the soft-clipping position is left-right, mark the SV type as repeat; and
[0056] If the reply direction is forward-reverse or reverse-forward, the SV type is marked as inverted;
[0057] SR signal clusters were formed by clustering SR signals marked as the same SV event according to SV type, reattachment direction and breakpoint coordinates.
[0058] Figure 2 An example of SV type judgment and labeling in the SR signal analysis process is shown in FIG.
[0059] In some embodiments of the above method, the PE signal analysis process may include:
[0060] The reads entering the PE signal analysis process are grouped according to QNAME (i.e., reads with the same QNAME are grouped into the same group), and in each group, R1 and R2 are paired according to the FLAG value of the reads, filtering out read pairs that cannot be paired;
[0061] Mark SV type, including:
[0062] In cases where R1 and R2 in the paired read pairs were located on different chromosomes, the SV type was marked as translocation;
[0063] In the case where R1 and R2 in the paired read length pair are located on the same chromosome, R1 and R2 are sorted according to the chromosome coordinates to obtain the re-posting direction information of the sorted read length pair, which has four possibilities: forward-reverse, reverse-forward, forward-forward and reverse-reverse;
[0064] If the reply direction is forward-reverse, the SV type is marked as missing;
[0065] If the reply direction is reverse-forward, the SV type is marked as duplicate;
[0066] If the reposting direction is forward-forward or reverse-reverse, the SV type is marked as inverted;
[0067] PE signal clusters were formed by clustering PE signals marked as the same SV event according to SV type, reattachment direction and breakpoint coordinates.
[0068] Figure 3 An example of SV type determination and labeling in the PE signal analysis process is shown in FIG.
[0069] In some embodiments of the above method, the method includes a step of filtering the read length when reading the sequence alignment file of the high-throughput sequencing data. In some embodiments, the read length whose MapQ is less than a set filtering threshold is filtered out when reading the sequence alignment file. In some embodiments, the filtering threshold can be set to an integer between 0 and 60. For example, the filtering threshold can be 0, 5, 10, 15 or 20.
[0070] In some embodiments, the read length located in the genomic repetitive region is filtered out when reading the sequence alignment file. It is known that there are a large number of repetitive regions in the human genome, including homopolymer sequences (repetition of a single base) and microsatellite sequences (repetition of 2-6 consecutive bases), etc., which seriously affect the reattachment. In some embodiments, the genomic repetitive region is defined as a region where the length of the homopolymer sequence or microsatellite sequence exceeds about 45bp.
[0071] In some embodiments, the SR signal analysis process may further include:
[0072] According to the starting coordinates and ending coordinates of the reads in the SR signal cluster, determine the number of reads (unique) and the total number of reads (total) containing the deduplicated template in each SR signal cluster;
[0073] Record the QNAME of the reads in the SR signal cluster whose number of reads of the template after deduplication is 1 and whose total number of reads is greater than the duplication threshold, and filter out the reads with the QNAME (the read pairs containing the reads) in the PE signal analysis process.
[0074] The repetition threshold is an integer between 2 and 10.
[0075] For example, the duplication threshold can be 2, 3, 4, 5, 6, 7, 8, 9, or 10. In some embodiments, the filtering threshold can be 3.
[0076] These QNAMEs (templates) may be from background noise or chimera false positives.
[0077] In some embodiments, the SR signal analysis process may further include merging SR signal clusters whose two coordinate fluctuations of the breakpoints are both less than or equal to a fluctuation range threshold, wherein the fluctuation range threshold is an integer between 1 bp and 100 bp.
[0078] For example, the fluctuation range threshold can be 1 bp, 2 bp, 5 bp, 10 bp, 20 bp, 30 bp, 40 bp, 50 bp, 60 bp, 70 bp, 80 bp, 90 bp or 100 bp. In some embodiments, the fluctuation range threshold can be 10 bp.
[0079] In some embodiments, the PE signal analysis process may further include merging PE signal clusters whose two coordinate fluctuations of the breakpoints are less than or equal to a fluctuation range threshold, wherein the fluctuation range threshold is an integer between 1 bp and 100 bp.
[0080] For example, the fluctuation range threshold can be 1 bp, 2 bp, 5 bp, 10 bp, 20 bp, 30 bp, 40 bp, 50 bp, 60 bp, 70 bp, 80 bp, 90 bp or 100 bp. In some embodiments, the fluctuation range threshold can be 10 bp.
[0081] In some embodiments of the above method, filtering the signal clusters may include performing similarity evaluation on the signal clusters and filtering out signal clusters with high similarity, wherein the evaluation includes:
[0082] Extract the reference genome sequences of 100-500 bp upstream and downstream of the two coordinates of the breakpoint and compare them. If the identity between the reference genome sequences upstream and downstream of the two coordinates of the breakpoint exceeds the identity threshold and the comparison length exceeds the comparison length threshold, the signal cluster is considered to have high similarity.
[0083] The identity threshold is a percentage between 70% and 100%, and the comparison length threshold is an integer between 40 bp and 100 bp.
[0084] In some embodiments, about 100 bp of reference genome sequence upstream and downstream of the two coordinates of the breakpoint are extracted and compared. In some embodiments, about 150 bp, about 200 bp, about 250 bp, about 300 bp, about 350 bp, about 400 bp, about 450 bp, or about 500 bp of reference genome sequence upstream and downstream of the two coordinates of the breakpoint are extracted and compared. In some embodiments, 200-250 bp, for example, about 220 bp of reference genome sequence upstream and downstream of the two coordinates of the breakpoint are extracted for comparison.
[0085] In some embodiments, the identity threshold can be, for example, 70%, 75%, 80%, 85%, 90%, 95%, 98% or 99%. For example, the identity threshold can be a percentage between 80%-90%.
[0086] In some embodiments, the comparison length threshold can be an integer between 45bp-55bp, such as 45bp, 46bp, 47bp, 48bp, 49bp, 50bp, 51bp, 52bp, 53bp, 54bp or 55bp.
[0087] In some embodiments, filtering the signal clusters comprises evaluating whether the signal clusters are located in genomic repetitive regions and filtering out signal clusters located in repetitive regions, wherein the evaluating comprises:
[0088] Extract the read sequences at the two coordinates of the signal cluster breakpoint, and identify whether the sequence contains homopolymer sequences and / or microsatellite sequences; if the read sequence at at least one coordinate contains homopolymer sequences and / or microsatellite sequences, it is considered that the signal cluster is located in the genomic duplication region,
[0089] The homopolymer sequence refers to a sequence in which one base is repeated 10 times or more, and the microsatellite sequence refers to a sequence in which 2-6 consecutive bases are repeated 4 times or more.
[0090] In some embodiments, filtering the signal clusters comprises evaluating whether the signal clusters have chimeras and filtering out signal clusters having chimeras, wherein the evaluating comprises:
[0091] Extract the soft-clipping sequence and the starting coordinates of the soft-clipping reads at each breakpoint coordinate, and group the reads with the same soft-clipping sequence and starting coordinates into one group;
[0092] The groups at each breakpoint coordinate are used to evaluate the statistical dispersion of the set formed by these groups.
[0093] In case the dispersion on at least one breakpoint coordinate is high, the signal cluster is considered to be mosaic.
[0094] In some embodiments, filtering the signal clusters comprises evaluating whether the signal clusters have chimeras and filtering out signal clusters having chimeras, wherein the evaluating comprises:
[0095] Extract the soft-clipping sequence and the starting coordinates of the soft-clipping reads at each breakpoint coordinate, and group the reads with the same soft-clipping sequence and starting coordinates into one group;
[0096] The group whose starting coordinates of soft-clipping are consistent with the breakpoint coordinates is taken as the main group, and the other groups are merged into secondary groups. The entropy values of the two groups are calculated.
[0097] In case the entropy value on at least one breakpoint coordinate exceeded 0.5, the signal cluster was considered to have chimeras.
[0098] In some embodiments, filtering the signal clusters comprises evaluating whether the signal clusters are simple templates and filtering out signal clusters that are simple templates, wherein the evaluating comprises:
[0099] Extract the fluctuation range of the start and end coordinates of the read template in the signal cluster. When the fluctuation range of the start and end coordinates of all read templates is less than or equal to the fluctuation range threshold, the signal cluster is considered to be a simple template.
[0100] The fluctuation range threshold is an integer between 2bp and 10bp.
[0101] For example, the fluctuation range threshold can be 2bp, 3bp, 4bp, 5bp, 6bp, 7bp, 8bp, 9bp or 10bp. In some embodiments, the fluctuation range threshold can be 3bp.
[0102] In some embodiments, filtering the signal clusters comprises evaluating the average mapq of the signal clusters and filtering out signal clusters with low average mapq, wherein the evaluating comprises:
[0103] Extract the mapq values of the reads at the two coordinates of the signal cluster breakpoint, and calculate the average mapq of the reads at each coordinate; if the average mapq of the reads at at least one coordinate is less than or equal to the filtering threshold, the signal cluster is considered to have a low average mapq.
[0104] The filtering threshold is an integer between 20 and 30.
[0105] For example, the filtering threshold can be 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, or 30. In some embodiments, the filtering threshold can be 25.
[0106] In some embodiments, filtering the signal clusters comprises evaluating whether the signal clusters exist in a BED file defined region, and filtering out signal clusters that are not in the BED file defined region, wherein the evaluating comprises:
[0107] Two coordinates of the signal cluster breakpoint are extracted. When both coordinates are not in the BED file definition area, it is considered that the signal cluster is not in the BED file definition area.
[0108] Positive detections that are not in the area defined in the BED file are generally not false positives, but rather areas of no concern or areas for which performance confirmation has not been performed.
[0109] In some embodiments, filtering the signal clusters includes filtering out signal clusters with SV sizes less than a certain threshold. For example, signal clusters with SV sizes less than 100 bp, 200 bp, 300 bp, 400 bp, 500 bp, 600 bp, 700 bp, 800 bp, 900 bp, or 1000 bp can be filtered out. In some embodiments, signal clusters with SV sizes less than 500 bp can be filtered out.
[0110] Positive detection of smaller SV size is generally not a false positive, but usually does not fall into the category of structural variation.
[0111] In some embodiments, filtering the signal clusters includes determining the number of read pairs (unique) and the total number of read pairs (total) containing deduplicated templates in each signal cluster based on the starting coordinates and ending coordinates of the reads in the signal clusters, and filtering out signal clusters whose number of read pairs of deduplicated templates is less than a filtering threshold.
[0112] In some embodiments, the filtering threshold can be an integer between 2 and 15. For example, the filtering threshold can be 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15.
[0113] In some embodiments, filtering the signal clusters may include any one or any combination of two or more of the above high similarity filtering, genomic repeat region filtering, chimera filtering, simple template filtering, low average mapq filtering, BED file defined region filtering, SV size filtering, and filtering based on the number of read pairs of different templates. For example, filtering the signal clusters may include one, two, three, four, five, six, seven, or all eight of the above filtering.
[0114] In some embodiments of the above method, the method may further comprise clustering and annotating SV abundances of the filtered signals.
[0115] In some embodiments, the method may further include annotating the filtered signal clusters with gene and exon / intron information at the two coordinate positions of the breakpoint, or annotating the gene information upstream and downstream thereof when the coordinates are at an intergenic position.
[0116] In some embodiments, the method is used to analyze SVs in a target genomic region, and the method includes filtering out reads that are not located in the target genomic region when reading a sequence alignment file of high-throughput sequencing data.
[0117] In terms of detection scope, the method disclosed in the present invention can not only detect gene or inter-gene structural variations, but also supports the detection of usually smaller intra-gene structural variations, which also have important clinical significance.
[0118] In terms of detection performance, the disclosed method improves sensitivity compared to traditional methods by independently extracting SR and PE signals and performing signal analysis processes separately, and finally optionally integrating the signals. At the same time, the accuracy can be greatly improved by developing a filtering system for sequencing data, genomic data, and biological characteristics.
[0119] In addition, the method disclosed herein supports a variety of sample types. The method disclosed herein can adapt to sample data with different abnormal insert length distributions by evaluating the abnormal insert length distribution, which are usually derived from different library construction methods. The method disclosed herein can also integrate the identification and filtering of chimeras, so that the method can analyze poor quality FFPE tissue samples with high sensitivity and high accuracy, which are often seen in clinical testing.
[0120] In one aspect, the present disclosure relates to a system for detecting structural variation in a genomic region, the system comprising:
[0121] High-throughput sequencing module; and
[0122] Structural variation detection module,
[0123] The high-throughput sequencing module is configured to perform high-throughput sequencing on the target genome region and generate a sequence alignment file; and the structural variation detection module is configured to detect structural variations in the target genome region by the method of the first aspect of the present disclosure.
[0124] In one aspect, the present disclosure relates to an apparatus for detecting structural variation in a genomic region, comprising:
[0125] memory for storing computer program instructions; and
[0126] a processor for executing computer program instructions,
[0127] Wherein, when the computer program instructions are executed by the processor, the device performs the method of the first aspect of the present disclosure.
[0128] In one aspect, the present disclosure relates to a computer-readable medium storing computer program instructions, wherein when the computer program instructions are executed by a processor, the method of the first aspect of the present disclosure is implemented. BRIEF DESCRIPTION OF THE DRAWINGS
[0129] Figure 1 A flow diagram of an exemplary method of the present disclosure is shown.
[0130] Figure 2 An example of SV type determination and labeling in the SR signal analysis process is shown in FIG.
[0131] Figure 3 An example of SV type determination and labeling in the PE signal analysis process is shown in FIG. DETAILED DESCRIPTION
[0132] Example 1. Structural variation detection method
[0133] In an embodiment of the present disclosure, the following method (markSV method) is used to detect structural variations in genomic regions.
[0134] 1. Generation of sequence alignment files: After the library preparation, the samples to be tested are sequenced on the Illumina sequencing platform to generate FASTQ files. After quality control of the FASTQ files, the alignment software BWA-MEM is used to align the FASTQ files with the human reference genome (hg19 / b37) and generate SAM files. The SAM files are converted to BAM files using the samtools software, and the BAM files are used as input files for subsequent testing.
[0135] 2. Calculation of insert length outliers: Read the reads in the BAM file to estimate the parameters of the insert length distribution and the threshold of the outlier. If the BAM file contains more than 1 million reads, 1 million are randomly selected for calculation to reduce the time of calculating the insert length distribution parameters. The insert length distribution defaults to log-normal distribution. The parameters μ and σ of the insert length distribution are calculated and the distribution outlier threshold is obtained according to the 3sigma principle. If the insert length of a pair of reads exceeds μ+3σ, it is considered an abnormal insert length.
[0136] 3. Read the BAM file, filter out reads with MapQ less than 10 and reads located in long repeat regions (homopolymer sequences or microsatellite sequences longer than 45 bp), and retain the remaining valid reads to proceed to the next step.
[0137] 4. Extract SV-related signals from valid reads: (1) Reads containing soft-clipping are treated as SR signals and enter the SR signal analysis process; (2) Reads whose reply direction is not forward-reverse are treated as PE signals and enter the PE signal analysis process; (3) Reads whose reply direction is forward-reverse but whose insert length exceeds the insert length distribution outlier threshold calculated in step 2 above are treated as PE signals and enter the PE signal analysis process.
[0138] 5.SR signal analysis process:
[0139] 5.1 The reads entering the SR signal analysis process are first grouped according to QNAME, and in each group, the primary and secondary replies are paired according to the FLAG value, and the reads that cannot be paired are filtered out;
[0140] 5.2 Read length of paired SR signal The SV type is determined based on the position and direction information of the primary and secondary replies, e.g. Figure 2 As shown in , SR signal clusters are formed after clustering according to SV type and breakpoint coordinate information;
[0141] 5.3 Count the number (unique) and total number (total) of reads containing different templates in the SR signal cluster according to the start and end coordinates of the read template. Record the QNAME of the reads in the SR signal cluster with unique = 1 and total greater than 3, which will be filtered out in the PE signal analysis process; and
[0142] 5.4 Cluster and merge the two SR signals whose two coordinate fluctuations of the breakpoints are less than or equal to 10 bp.
[0143] 6.PE signal analysis process:
[0144] 6.1 The read pairs entering the PE signal analysis process are first grouped according to QNAME, and R1 and R2 are paired according to the FLAG value in each group to filter out the reads that cannot be paired;
[0145] 6.2 The paired PE signal read length pair determines the SV type based on the position and direction information of R1 and R2, such as Figure 3 As shown in ; filter out the reads with the QNAME recorded in the above step 5.4, and form PE signal clusters after clustering according to SV type and breakpoint coordinate information;
[0146] 6.3 Cluster and merge the two PE signals whose two coordinate fluctuations of the breakpoints are both less than or equal to 10bp.
[0147] 7 Signal integration: SR signal clusters and PE signal clusters are merged according to the breakpoint coordinate information and the QNAME of the read length to form a comprehensive signal cluster, which includes the merged SR signal clusters and PE signal clusters, or the individual SR signal clusters or PE signal clusters that failed to be merged.
[0148] 8. Filters:
[0149] 8.1 Evaluate the similarity of the integrated signal clusters and filter out the integrated signal clusters with high similarity, as follows: extract the reference genome sequence of 220 bp upstream and downstream of the two coordinates of the breakpoint, and use BLASTn to compare them. If the identity exceeds 80% and the comparison length exceeds 50 bp, it is considered to be highly similar;
[0150] 8.2 Evaluate the repetitive regions for the integrated signal clusters and filter out the integrated signal clusters located in the repetitive regions, as follows: extract the read sequences at the two coordinates of the breakpoints of the integrated signal clusters, and identify whether the sequences contain homopolymers (1 base sequence repeated more than 10 times), microsatellites (2-6 consecutive base sequences repeated more than 4 times), or a combination of the two; if the read sequence at at least one coordinate contains homopolymer sequences and / or microsatellite sequences, it is considered to be located in the repetitive region;
[0151] 8.3 Evaluate the chimeras for the integrated signal clusters and filter out the integrated signal clusters with chimeras, as follows: extract the soft-clipping sequence and the starting coordinates of the soft-clipping reads at each breakpoint coordinate, and group the reads with both the soft-clipping sequence and the starting coordinates of the soft-clipping into one group; take the group with the starting coordinates of the soft-clipping consistent with the breakpoint coordinates as the main group, merge the other groups into the secondary group, and calculate the entropy values of the two groups; if the entropy value on at least one breakpoint coordinate exceeds 0.5, the integrated signal cluster is considered to have chimeras.
[0152] 8.4 Evaluate the simple templates for the comprehensive signal clusters and filter out the comprehensive signal clusters that are simple templates, as follows: extract the fluctuation range of the start and end coordinates of the read template in the comprehensive signal clusters. If the fluctuation range of the start and end coordinates is less than or equal to 3 bp, it is considered to be a simple template;
[0153] 8.5 Evaluate the average mapq of the integrated signal clusters and filter out the integrated signal clusters with low average mapq, specifically as follows: extract the mapq values of the reads at the two coordinates of the breakpoints of the integrated signal clusters, and calculate the average mapq; if the average mapq of the reads at at least one coordinate is less than or equal to the filtering threshold of 25, the integrated signal cluster is considered to have a low average mapq;
[0154] 8.6 Evaluate whether the integrated signal cluster is in the BED file definition area and filter out the integrated signal clusters that are not in the BED file definition area, specifically as follows: extract two coordinates of the integrated signal cluster breakpoints, and if both coordinates are not in the BED file definition area, it is considered that the integrated signal cluster is not in the BED file definition area; if no BED file is specified, this step is not performed;
[0155] 8.7 Filter out the integrated signal clusters with SV size less than 500;
[0156] 8.8 According to the starting coordinates and ending coordinates of the reads in the integrated signal clusters, determine the number of read pairs (unique) and the total number of read pairs (total) containing different templates in each integrated signal cluster, and filter out integrated signal clusters with unique values less than 4.
[0157] 9. Output report: Annotate the comprehensive signal clustering and output the result report.
[0158] Example 2. Simulation data performance confirmation
[0159] The simulated data covers 79 common fusion genes, a total of 11.81M regions. Various SV types were simulated, and a total of 20,000 SVs ranging in size from 50bp to 1,000,000bp were included. The sequencing depth was set at 200x, and the SV abundance was set at 50%. The method described in Example 1 (markSV method) was used for comparison with three other mainstream SV analysis software. These three SV analysis software are Delly v0.7.9, Lumpy v0.2.13, and Manta v1.4.0. All were analyzed with default parameters. The results of FILTER being PASS in the VCF file were retained, where Lumpy did not set FILTER, and all results were retained. The analysis results are as follows:
[0160] Table 1. Sensitivity and accuracy of markSV in 20,000 broad-spectrum simulated data
[0161]
[0162] Table 2. Sensitivity and precision of four SV analysis software in 20,000 broad spectrum simulation data
[0163] SV Analysis Software Sensitivity Accuracy markSV 96.89% 99.94% DELLY 87.79% 45.54% LUMPY 95.73% 48.93% Manta 96.49% 86.40%
[0164] The results show that the method disclosed in the present invention has a sensitivity (PPA, positive agreement rate) of 96.89% and an accuracy (PPV, positive predictive value) of 99.94% in simulated data. It has the best sensitivity and accuracy among the four methods, and has a greater advantage in accuracy than the other three methods.
[0165] Example 3. Standard product data
[0166] The standard data was prepared by gradient dilution of a mixed sample of HD-C670 cell line containing two fusions (EML4-ALK, CD74-ROS1), and the abundance after dilution was calibrated by the ddPCR platform. To ensure reproducibility, each gradient was repeated multiple times under different machine batches, different reagent batches, different experimental operators, etc., for a total of 58 samples. The results are as follows:
[0167] Table 3. Summary of Standard Data
[0168]
[0169] The method described in Example 1 (markSV method) was used for comparison with three other mainstream SV analysis software. The three SV analysis software were Delly v0.7.9, Lumpy v0.2.13, and Manta v1.4.0. All were analyzed with default analysis parameters. The results of FILTER PASS in the VCF file were retained, and Lumpy did not set FILTER, and all results were retained. The analysis results are as follows:
[0170] Table 4. Sensitivity and precision of four SV analysis software in standard sample data
[0171] SV Analysis Software True Yang Fake vagina False Penis Sensitivity Accuracy markSV 232 0 31 100% 88.21% DELLY 230 2 941 97.34% 19.64% LUMPY 174 58 3963 96.62% 4.21% Manta 232 0 624 100% 27.1%
[0172] For the 58 samples of the standard, since they contain two fusions, EML4-ALK and CD74-ROS1, and both fusions are balanced translocation events, each sample contains four fusions, namely EML4-ALK, ALK-EML4, CD74-ROS1, and ROS1-CD74. In summary, for the standard data, there are a total of 4*58=232 true positive results (TP).
[0173] The results also showed that the sensitivity (PPA, positive coincidence rate) of this method in the simulated data was 100%, and the accuracy (PPV, positive predictive value) was 88.21%. Compared with the other three methods, this method has the best sensitivity and accuracy, and has a greater advantage in accuracy.
Claims
1. A method for detecting a structural variation (SV) in a genomic region, the method comprising: a. Read the sequence alignment file of high-throughput sequencing data, i) The read length containing soft-clipping is used as the SR (Split-read) signal to enter the SR signal analysis process; and / or ii) read pairs whose re-attachment direction in the reference genome is not forward-reverse, and read pairs whose re-attachment direction in the reference genome is forward-reverse but with abnormal insert length (insert size) are entered into the PE signal analysis process as PE (Paired-end) signals; b. The reads entering the SR signal analysis process are grouped according to QNAME, and the main replies and suboptimal replies are paired according to the FLAG value of the reads in each group, and the reads that cannot be paired are filtered out; the SV type is determined according to the position and direction information of the main replies and suboptimal replies of the paired SR signal reads, and the SR signals marked as the same SV event are clustered according to the SV type, the replies direction and the breakpoint coordinates to form SR signal clusters; and / or the reads entering the PE signal analysis process are grouped according to QNAME, and R1 and R2 are paired in each group according to the FLAG value of the reads, and the read pairs that cannot be paired are filtered out; the SV type is determined based on the position and direction information of the paired R1 and R2, and the PE signals marked as the same SV event are clustered according to the SV type, the replies direction and the breakpoint coordinates to form PE signal clusters; c. filtering the SR signal cluster and / or the PE signal cluster to obtain a filtered signal cluster indicating an SV in a genomic region; wherein the filtered signal clusters include at least one of the following: a signal cluster with high similarity, a signal cluster located in a genomic repeat region, a chimera signal cluster, a signal cluster of a simple template, a signal cluster of a low average mapq, a signal cluster not in a BED file defined region, a signal cluster with an SV size less than a certain threshold, and a signal cluster with a number of read length pairs of a template after deduplication less than a filtering threshold; and optionally d. Output result report.
2. A method for detecting a structural variation (SV) in a genomic region, the method comprising: a. Read the sequence alignment file of high-throughput sequencing data, and enter the SR signal analysis process with the read length containing soft-clipping as the SR signal; enter the PE signal analysis process with the read length pairs whose re-attachment direction in the reference genome is not forward-reverse, and the read length pairs whose re-attachment direction in the reference genome is forward-reverse but with abnormal insertion fragment length as PE signals; b. The reads entering the SR signal analysis process are grouped according to QNAME, and the main post-post and the suboptimal post-post are paired according to the FLAG value of the read in each group, and the reads that cannot be paired are filtered out; the SV type is determined according to the position and direction information of the main post-post and the suboptimal post-post of the paired SR signal reads, and the SR signals marked as the same SV event are clustered according to the SV type, post-post direction and breakpoint coordinates to form SR signal clusters; and, the reads entering the PE signal analysis process are grouped according to QNAME, and R1 and R2 are paired in each group according to the FLAG value of the read, and the read pairs that cannot be paired are filtered out; the SV type is determined based on the position and direction information of the paired R1 and R2, and the PE signals marked as the same SV event are clustered according to the SV type, post-post direction and breakpoint coordinates to form PE signal clusters; c. merging the SR signal cluster and the PE signal cluster marked as the same SV event according to the breakpoint coordinate information and the QNAME of the read length to form a comprehensive signal cluster, wherein the comprehensive signal cluster includes the merged SR signal cluster and the PE signal cluster, or the separate SR signal cluster or the PE signal cluster that failed to be merged; d. filtering the comprehensive signal clusters to obtain filtered comprehensive signal clusters, which indicate SVs in genomic regions; wherein the filtered comprehensive signal clusters include any one of the following: comprehensive signal clusters with high similarity, comprehensive signal clusters located in repeated regions, chimera comprehensive signal clusters, comprehensive signal clusters of simple templates, comprehensive signal clusters of low average mapq, comprehensive signal clusters not in the BED file defined region, comprehensive signal clusters with SV size less than a certain threshold, and comprehensive signal clusters with a number of read pairs less than a filtering threshold; and optionally e. Output result report.
3. The method of claim 1 or 2, wherein the sequence alignment file is selected from a SAM file and a BAM file.
4. The method of claim 1 or 2, wherein the read length with abnormal insert length is determined by the following steps: Read the insert length values of all reads in the sequence alignment file, or randomly read the insert length values of a certain number of reads; constructing an insert length distribution using the insert length values, and determining an outlier threshold value significantly above the insert length distribution; and Reads having insert lengths exceeding the outlier threshold are determined.
5. The method of claim 4, wherein the read length with abnormal insert length is determined by the following steps: constructing a normal or log-normal distribution using the insert length values, and calculating parameters μ and σ of the normal or log-normal distribution; and The insert fragment length value exceeding μ+Nσ is determined as having an abnormal insert fragment length value, Wherein N is an integer between 3 and 10.
6. The method according to claim 1 or 2, wherein the paired SR signal read length determines the SV type according to the position and direction information of the primary reply and the secondary reply, comprising: In cases where the paired primary and secondary reattachments were located on different chromosomes, the SV type was marked as a translocation; In the case where the paired primary reattachment and the suboptimal reattachment are located on the same chromosome, the primary reattachment and the suboptimal reattachment are sorted according to the chromosome coordinates to obtain the sorted reattachment direction information; and at the same time, the soft-clipping position information of the sorted primary reattachment and the suboptimal reattachment is obtained; If the reclipping direction is forward-forward or reverse-reverse, and the soft-clipping position is right-left, the SV type is marked as missing; If the reclipping direction is forward-forward or reverse-reverse, and the soft-clipping position is left-right, mark the SV type as repeat; and If the reply direction is forward-reverse or reverse-forward, the SV type is marked as inverted; SR signal clusters were formed by clustering SR signals marked as the same SV event according to SV type, reattachment direction and breakpoint coordinates.
7. The method according to claim 1 or 2, wherein determining the SV type based on the position and direction information of the paired R1 and R2 comprises: In cases where R1 and R2 in the paired read pairs were located on different chromosomes, the SV type was marked as translocation; When R1 and R2 in the paired read length pair are located on the same chromosome, R1 and R2 are sorted according to the chromosome coordinates to obtain the re-attachment direction information of the sorted read length pair; If the reply direction is forward-reverse, the SV type is marked as missing; If the reply direction is reverse-forward, the SV type is marked as duplicate; If the reposting direction is forward-forward or reverse-reverse, the SV type is marked as inverted; PE signal clusters were formed by clustering PE signals marked as the same SV event according to SV type, reattachment direction and breakpoint coordinates.
8. The method of claim 1 or 2, wherein when reading the sequence alignment file, reads with MapQ less than a filtering threshold are filtered out, The filtering threshold is an integer between 0 and 60.
9. The method of claim 8, wherein the filtering threshold is selected from 0, 5, 10, 15 and 20.
10. The method according to claim 1 or 2, wherein the read length located in the genomic repetitive region is filtered out when reading the sequence alignment file, and the genomic repetitive region is a region where the repetition length of 1 base or 2-6 consecutive bases exceeds 45bp.
11. The method according to claim 1 or 2, wherein the SR signal analysis process further comprises: According to the starting coordinates and ending coordinates of the reads in the SR signal cluster, determine the number of reads (unique) and the total number of reads (total) containing the deduplicated template in each SR signal cluster; Record the QNAME of the reads in the SR signal cluster whose number of reads of the template after deduplication is 1 and whose total number of reads is greater than the duplication threshold, and filter out the reads with the QNAME in the PE signal analysis process. The repetition threshold is an integer between 2 and 10. The method of claim 11 , wherein the repetition threshold is 3.
13. The method according to claim 1 or 2, wherein the SR signal analysis process further comprises merging SR signal clusters whose two coordinate fluctuations of the breakpoints are both less than or equal to a fluctuation range threshold, wherein the fluctuation range threshold is an integer between 1 bp and 100 bp. The method according to claim 13 , wherein the fluctuation range threshold is 10 bp.
15. The method according to claim 1 or 2, wherein the PE signal analysis process further comprises merging PE signal clusters whose two coordinate fluctuations of the breakpoints are both less than or equal to a fluctuation range threshold, wherein the fluctuation range threshold is an integer between 1 bp and 100 bp. The method according to claim 15 , wherein the fluctuation range threshold is 10 bp.
17. The method of claim 1, wherein filtering the signal clusters comprises performing similarity evaluation on the signal clusters and filtering out signal clusters with high similarity, wherein the evaluation comprises: Extract the reference genome sequences of 100-500 bp upstream and downstream of the two coordinates of the breakpoint and compare them. If the identity between the reference genome sequences upstream and downstream of the two coordinates of the breakpoint exceeds the identity threshold and the comparison length exceeds the comparison length threshold, the signal cluster is considered to have high similarity. The identity threshold is a percentage between 70% and 100%, and the comparison length threshold is an integer between 40 bp and 100 bp.
18. The method of claim 17, wherein 200-250 bp of reference genome sequence upstream and downstream of the two coordinates of the breakpoint are extracted for comparison.
19. The method of claim 17 or 18, wherein the identity threshold is a percentage between 80% and 90%, and the comparison length threshold is an integer between 45 bp and 55 bp.
20. The method of claim 1, wherein filtering the signal clusters comprises evaluating whether the signal clusters are located in genomic repetitive regions and filtering out signal clusters located in repetitive regions, wherein the evaluating comprises: Extract the read sequences at the two coordinates of the signal cluster breakpoint, and identify whether the sequence contains homopolymer sequences and / or microsatellite sequences; if the read sequence at at least one coordinate contains homopolymer sequences and / or microsatellite sequences, it is considered that the signal cluster is located in the genomic duplication region, The homopolymer sequence refers to a sequence in which one base is repeated 10 times or more, and the microsatellite sequence refers to a sequence in which 2-6 consecutive bases are repeated 4 times or more.
21. The method of claim 1, wherein filtering the signal clusters comprises evaluating whether the signal clusters are chimeras and filtering out signal clusters that are chimeras, wherein the evaluating comprises: Extract the soft-clipping sequence and the starting coordinates of the soft-clipping reads at each breakpoint coordinate, and group the reads with the same soft-clipping sequence and starting coordinates into one group; Using the groups at each breakpoint coordinate, evaluating the degree of dispersion (statistical dispersion) of the set formed by these groups; In case the dispersion on at least one breakpoint coordinate is high, the signal cluster is considered to be mosaic.
22. The method of claim 21, wherein determining whether the signal cluster is a chimera is performed by: The group whose starting coordinates of soft-clipping are consistent with the breakpoint coordinates is taken as the main group, and the other groups are merged into secondary groups. The entropy values of the two groups are calculated. In case the entropy value on at least one breakpoint coordinate exceeded 0.5, the signal was considered clustered as a chimera.
23. The method of claim 1, wherein filtering the signal clusters comprises evaluating whether the signal clusters are simple templates and filtering out signal clusters that are simple templates, wherein the evaluating comprises: Extract the fluctuation range of the start and end coordinates of the read template in the signal cluster. When the fluctuation range of the start and end coordinates of all read templates is less than or equal to the fluctuation range threshold, the signal cluster is considered to be a simple template. The fluctuation range threshold is an integer between 2bp and 10bp.
24. The method of claim 23, wherein the fluctuation range threshold is 3 bp.
25. The method of claim 1, wherein filtering the signal clusters comprises evaluating average mapq of the signal clusters and filtering out signal clusters with low average mapq, the evaluating comprising: Extract the mapq values of the reads at the two coordinates of the signal cluster breakpoints, and calculate the average mapq of the reads at each coordinate; If the average mapq of the reads on at least one coordinate is less than or equal to the filtering threshold, the signal cluster is considered to have a low average mapq. The filtering threshold is an integer between 20 and 30. The method of claim 25 , wherein the filtering threshold is 25.
27. The method of claim 1, wherein filtering the signal clusters comprises determining the number of read pairs (unique) and the total number of read pairs (total) containing deduplicated templates in each signal cluster according to the start coordinates and the end coordinates of the reads in the signal clusters, and filtering out signal clusters whose number of read pairs of deduplicated templates is less than a filtering threshold, The filtering threshold is an integer between 2 and 15.
28. The method of claim 27, wherein the filtering threshold is 2, 3, 4, 5, 6 or 7.
29. The method of claim 1, wherein the method further comprises annotating the gene and exon / intron information at the two coordinate positions of the breakpoint for the filtered signal cluster, or annotating the gene information upstream and downstream thereof when the coordinate is at an intergenic position.
30. The method of claim 2, wherein filtering the integrated signal clusters comprises performing similarity evaluation on the integrated signal clusters and filtering out integrated signal clusters with high similarity, wherein the evaluation comprises: Extract 220bp of reference genome sequences upstream and downstream of the two coordinates of the breakpoint and compare them. If the identity between the reference genome sequences upstream and downstream of the two coordinates of the breakpoint exceeds the identity threshold and the comparison length exceeds the comparison length threshold, the comprehensive signal clustering is considered to have high similarity. The identity threshold is 80%, and the comparison length threshold is 50 bp.
31. The method of claim 2, wherein filtering the integrated signal clusters comprises evaluating whether the integrated signal clusters are located in genomic repetitive regions and filtering out signal clusters located in repetitive regions, wherein the evaluating comprises: Extracting read sequences at two coordinates of the integrated signal cluster breakpoints, and identifying whether the sequences contain homopolymer sequences and / or microsatellites; if the read sequences at at least one coordinate contain homopolymer sequences and / or microsatellite sequences, it is considered that the integrated signal cluster is located in a genomic duplication region; The homopolymer sequence refers to a sequence in which one base is repeated 10 times or more, and the microsatellite sequence refers to a sequence in which 2-6 consecutive bases are repeated 4 times or more.
32. The method of claim 2, wherein filtering the integrated signal clusters comprises evaluating whether the integrated signal clusters are chimeras and filtering out integrated signal clusters that are chimeras, wherein the evaluating comprises: Extract the soft-clipping sequence and the starting coordinates of the soft-clipping reads at each breakpoint coordinate, and group the reads with the same soft-clipping sequence and starting coordinates into one group; Using the groups at each breakpoint coordinate, evaluating the degree of dispersion (statistical dispersion) of the set formed by these groups; In case the dispersion on at least one breakpoint coordinate is high, the integrated signal cluster is considered to be a mosaic.
33. The method of claim 32, wherein determining whether the integrated signal cluster is a mosaic is performed by: The group whose starting coordinates of softclipping are consistent with the breakpoint coordinates is taken as the main group, and the other groups are merged into the secondary group, and the entropy values of the two groups are calculated; In cases where the entropy value on at least one breakpoint coordinate exceeded 0.5, the integrated signal cluster was considered to have a chimera.
34. The method of claim 2, wherein filtering the integrated signal clusters comprises evaluating whether the integrated signal clusters are simple templates and filtering out integrated signal clusters that are simple templates, wherein the evaluating comprises: The fluctuation range of the start and end coordinates of the read template in the comprehensive signal cluster is extracted. When the fluctuation range of the start and end coordinates of all read templates is less than or equal to the fluctuation range threshold, the comprehensive signal cluster is considered to be a simple template.
35. The method of claim 34, wherein the fluctuation range threshold is 3 bp.
36. The method of claim 2, wherein filtering the integrated signal clusters comprises evaluating average mapq of the integrated signal clusters and filtering out integrated signal clusters with low average mapq, wherein the evaluating comprises: Extract the MAPQ values of the read lengths at the two coordinates of the integrated signal clustering breakpoints and calculate the average MAPQ; In the case where the average mapq of the reads at at least one coordinate is less than or equal to the filtering threshold, the integrated signal cluster is considered to be low average mapq. The method of claim 36 , wherein the filtering threshold is 25.
38. The method of claim 2, wherein filtering the integrated signal clusters comprises evaluating whether the integrated signal clusters are not in a BED file defined area and filtering out the integrated signal clusters not in the BED file defined area, the evaluating comprising: Two coordinates of the breakpoint of the integrated signal cluster are extracted. When both coordinates are not in the BED file definition area, it is considered that the integrated signal cluster is not in the BED file definition area.
39. The method of claim 2, wherein filtering the integrated signal clusters comprises filtering out integrated signal clusters whose SV sizes are smaller than a certain threshold.
40. The method of claim 39, wherein the certain threshold is 200 bp.
41. The method of claim 2, wherein filtering the integrated signal clustering comprises determining the number of read pairs (unique) and the total number of read pairs (total) containing deduplicated templates in each integrated signal clustering according to the starting coordinates and the ending coordinates of the reads in the integrated signal clustering, and filtering out signal clusters whose number of read pairs of the deduplicated templates is less than a filtering threshold.
42. The method of claim 41, wherein the filtering threshold is 4.
43. The method of claim 1 or 2, wherein the method is used to analyze SVs in a target genomic region, and the method comprises filtering out reads that are not located in the target genomic region when reading a sequence alignment file of high-throughput sequencing data.
44. A system for detecting structural variation in a genomic region, the system comprising: High-throughput sequencing module; and Structural variation detection module, The high-throughput sequencing module is configured to perform high-throughput sequencing on the target genomic region and generate a sequence alignment file; and the structural variation detection module is configured to detect structural variations in the target genomic region by the method described in any one of claims 1-43.
45. An apparatus for detecting structural variation in a genomic region, comprising: memory for storing computer program instructions; and a processor for executing computer program instructions, Wherein when the computer program instructions are executed by the processor, the device performs the method of any one of claims 1-43.
46. A computer-readable medium storing computer program instructions, wherein when the computer program instructions are executed by a processor, the method of any one of claims 1-43 is implemented.
