A system and method for structural variant validation of genomic repeat sequences

Abstract

This invention discloses a system and method for verifying structural variations in genomic repetitive sequences, comprising: a sequence alignment module for aligning sequences in a sorted BAM file, a reference genome's FASTA file, and a VCF file to obtain re-alignment results, which are then sent to a deduplication module; a deduplication module for deduplicating the received re-alignment results and sending the deduplicated results to a structural variation evaluation module; and a structural variation evaluation module for evaluating the received deduplicated results using a distance-based structural variation evaluation method, thereby verifying structural variations in genomic repetitive sequences. This structural variation verification system helps people efficiently, accurately, and comprehensively examine sequencing evidence of structural variations, enhancing and simplifying the manual review process.

Description

Technical Field

[0001] This invention belongs to the field of genome structural variation identification technology, specifically relating to a system and method for verifying structural variations in genome repetitive sequences. Background Technology

[0002] Since the discovery of the DNA double helix structure, life science research has entered the molecular level. Sequencing technology, which emerged in the 1970s, has made a significant contribution to deciphering the genetic code. Single-molecule sequencing technology, which has emerged in recent years, can read nucleotide sequences at the single-molecule level and is also known as third-generation sequencing technology. Major examples include the Pacific Bioscience (PacBio) and Oxford Nanopore Technology (ONT) platforms. Compared to traditional first- and second-generation sequencing technologies, third-generation sequencing can produce longer base reads, directly sequence RNA without reverse transcription, and is extremely fast. Furthermore, single-molecule sequencing technology provides the opportunity to detect structural variations at higher resolution and more comprehensively. The average sequence length of third-generation single-molecule sequencing is 15 kbp or higher, thus significantly improving the reliability and resolution of structural variation detection, especially in repetitive regions and complex structural variations in the human genome.

[0003] Genomic structural variation refers to genomic rearrangements exceeding 50 bp in length, typically including deletions, insertions, inversions, duplications, and translocations. Structural variations are relevant to everyone, manifesting primarily in human diseases (cancer, autism, Alzheimer's disease, etc.), chromosomal evolution (gene loss and transposon activity), gene regulation (transcription factor rearrangements), and other phenotypes (mating and intrinsic reproductive isolation). Therefore, the characteristics of structural variations are of great significance to human medicine and genetics. They aid in the early detection of diseases and help elucidate their underlying genetic and molecular processes.

[0004] Accurate identification of structural variations is a prominent yet crucial issue in genomics. The rapid development of single-molecule sequencing technology has provided better resolution and more comprehensive detection opportunities for whole-genome structural variations. In recent years, numerous studies and tools based on single-molecule sequencing for genome structural variation detection have emerged, such as Sniffles (published in 2018), Svim (published in 2019), and CuteSV (published in 2020). However, even the most advanced tools still produce a large number of false positives, thus necessitating validation and evaluation of the results of structural variation detection.

[0005] Currently, there are limited studies on the validation and assessment of structural variations, with Vapor and TT-Mars being the main available methods. However, both tools are ineffective at validating and assessing structural variations within repetitive sequence regions of the genome, and their visualization capabilities are relatively poor. Furthermore, TT-Mars results rely excessively on high-quality genome assembly sequences. Therefore, there are currently no effective tools for validating structural variations within repetitive sequence regions of the genome. Summary of the Invention

[0006] To overcome the problems in the prior art, the purpose of this invention is to provide a system and method for verifying structural variations in genomic repetitive sequences, which can accurately assess structural variations in genomic repetitive sequences.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A structural variation verification system for genomic repetitive sequences includes:

[0009] The sequence alignment module is used to perform sequence alignment on the sorted BAM file, the reference genome's FASTA file, and the VCF file using a kmer-based hash alignment algorithm. It obtains the re-alignment results of the sequencing read containing structural variations with the reference genome's ref sequence and the structural variation prediction sequence pre, and sends them to the deduplication module.

[0010] The deduplication module is used to deduplicatively compare the received sequencing read containing structural variations with the reference genome ref sequence and the structural variation prediction sequence pre, and then send the deduplication result to the structural variation evaluation module.

[0011] The structural variation assessment module is used to evaluate the received deduplication results using a distance-based structural variation assessment method, thereby verifying the structural variations of genomic repetitive sequences.

[0012] A method for validating structural variations in genomic repetitive sequences includes the following steps:

[0013] 1) For the sorted BAM file, the reference genome's FASTA file and VCF file, sequence alignment is performed using a kmer-based hash alignment algorithm to obtain the re-alignment results of the sequencing read containing structural variations with the reference genome's ref sequence and the structural variation prediction sequence pre.

[0014] 2) Remove duplicates from the alignment results of the sequencing reads containing structural variations with the ref sequences of the reference genome and the predicted structural variation sequences pre;

[0015] 3) The deduplication results are evaluated using a distance-based structural variation assessment method to verify the structural variations of genomic repetitive sequences.

[0016] Furthermore, the specific process of step 1) is as follows:

[0017] Extract the start and end coordinates of each structural variation within the VCF file;

[0018] Based on the start and end coordinates of the structural variant, all m sequencing reads covered by a certain structural variant s and the corresponding reference genome sequence ref are retrieved from the BAM file. The structural variant s is then inserted into the reference genome sequence ref to construct a structural variant prediction sequence pre.

[0019] Then, the kmer sequence is traversed on the reference genome sequence ref. For each new kmer sequence obtained, the hash table is indexed to see if there is a matching kmer sequence. If there is, the kmer sequence position that matches in the sequencing read and the reference genome sequence ref is moved by 1 bp to check if the next base is the same, until the bases of the two are different. The starting coordinates of the matching sequence in the sequencing read and the reference genome sequence ref, the length of the matching sequence, and the direction of the matching sequence are recorded. The re-alignment results of the sequencing read containing structural variations, the reference genome ref sequence, and the structural variation prediction sequence pre are obtained.

[0020] Furthermore, the hash table is determined through the following process:

[0021] The sequencing reads were re-aligned with their corresponding reference genome sequences (ref) and structural variation prediction sequences (pre). For the sequencing data of CCS and ONT, the length of the kmer sequence was selected as 31. The sequencing reads were traversed, and the starting coordinates of the kmer sequences were selected as the hash values ​​of the kmer sequences and stored in a hash table. The reverse complementary sequences of the sequencing reads were traversed, and the kmer sequences and their corresponding starting coordinates were stored in a hash table.

[0022] Furthermore, the specific process of step 2) is as follows:

[0023] If there are no repetitive sequences in the reference genome ref sequence, the result of the re-alignment is a line segment with a length equal to the length of the reference genome ref sequence in a Cartesian coordinate system, and is represented as a line segment along the main diagonal in the sequence alignment diagram;

[0024] If the reference genome ref sequence contains repetitive sequences, the re-alignment result includes several line segments of varying lengths, which are represented in the sequence alignment map as several line segments not along the main diagonal, recording the coordinate range of the ref of the repetitive sequence.

[0025] Furthermore, step 2) also includes the following steps:

[0026] The results of the re-alignment of the sequencing read with the reference genome's ref sequence are traversed according to the ref coordinate order of the repetitive sequences. In the sequence alignment diagram, for fragments located to the left of the structural variant, duplicates are removed while fragments located on the main diagonal are retained; for fragments located to the right of the structural variant, duplicates are removed while fragments located on the line that is offset from the main diagonal by the length intercept of the structural variant are retained.

[0027] Furthermore, the specific process of step 3) is as follows:

[0028] The fragments generated by re-aligning the sequencing read with the reference genome sequence ref and the fragments generated by re-aligning the sequencing read with the structural variant prediction sequence pre are visualized to generate sequence alignment maps. The fragments generated by re-alignment are represented as several line segments in the sequence alignment map. If the sequencing read is completely identical to the reference genome sequence ref, there will be a line segment in the generated sequence alignment map that runs exactly along the main diagonal. Otherwise, the line segment in the sequence alignment map will deviate from the main diagonal, and the magnitude of the deviation reflects the degree of difference between the two sequences. By calculating the average distance between all line segments in the sequence alignment map and the main diagonal, the average distance in the sequence alignment map corresponding to the structural variant s, the average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the structural variant prediction sequence pre are obtained. The average distances between the sequencing read and the reference genome sequence ref and the average distances between the sequencing read and the structural variant prediction sequence pre are normalized to obtain the score of the structural variant s. The structural variant is evaluated based on the score of the structural variant s.

[0029] Furthermore, the average distance d in the sequence alignment plot corresponding to the structural variation s s,i,j,avg Calculated using the following formula:

[0030]

[0031] Suppose a structural variant *s* covers *m* sequencing reads, where the sequence alignment graph corresponding to the *i*th sequencing read has *n* line segments, and the distance *d* of the *j*th line segment from the main diagonal is... s,i,j Calculated using the following formula:

[0032] d s,i,j =1 / 3[(xs,i,j,start -y s,i,j,start )+(x s,i,j,mid -y s,i,j,mid )+(x s,i,j,end -y s,i,j,end )]

[0033] In the formula, x s,i,j,start y s,i,j,start Let x be the starting coordinate of the ref corresponding to the j-th line segment. s,i,j,mid y s,i,j,mid Let x be the coordinates of the midpoint of the j-th line segment, x is the midpoint of the read line segment. s,i,j,end y s,i,j,end Let ref be the coordinates of the endpoint of the j-th line segment, and let read be the coordinates of the endpoint of the line segment.

[0034] Furthermore, the average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the predicted structural variant sequence pre, are normalized to obtain the score for the structural variant s. s,i :

[0035]

[0036] When the score of structural variation s is greater than the threshold Score threshold At that time, the structural variation s is considered to be a qualified structural variation for evaluation.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention first performs k-mer hash-based re-alignment of the sequencing reads within each structural variant region with two reference genome sequences. Second, the re-alignment results are deduplicated to eliminate the influence of repetitive sequences in the reference genome on structural variant assessment. Third, based on the deduplicated re-alignment results, a distance-based structural variant assessment model is used to score and validate each structural variant. Finally, the assessment results of each structural variant and a visualization image are output together. Experimental testing shows that this structural variant verification system can effectively assess and verify structural variants with repetitive sequences in the genome, and the output visualization images allow users to more intuitively view the true situation of each structural variant. Therefore, this structural variant verification system helps people efficiently, accurately, and comprehensively review sequencing evidence of structural variants, enhancing and simplifying the manual review process. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of a structural variation verification system for genomic repetitive sequences according to the present invention;

[0040] Figure 2 This is a predicted sequence constructed using deletion as an example;

[0041] Figure 3 This is a schematic diagram illustrating the principle of deduplication.

[0042] Figure 4 ROC curves for 21,000 simulated data points;

[0043] Figure 5 A graph showing the recall rate of 21,000 simulated data points as a function of a scoring threshold;

[0044] Figure 6 The diagram shows the change in recall rate of structural variations in repetitive regions with the scoring threshold; where (a) is the result diagram of repetitive and non-repetitive regions in homozygous sequences at different sequencing depths, and (b) is the result diagram of repetitive and non-repetitive regions in homozygous and heterozygous sequences at the same sequencing depth.

[0045] Figure 7 The graph shows the changes in recall rates of spotsv and vapor as a function of the scoring threshold in the HG002 sample.

[0046] Figure 8 Taking deletion as an example, the input and output of the structural variation verification system are as follows: (a) is the segment plot of the reference genome sequence (ref) and its own sequence alignment plot, (b) is the segment plot of the pre-deduplication sequencing read and the reference genome sequence (ref), and (c) is the segment plot of the post-deduplication sequencing read and the reference genome sequence (ref). Detailed Implementation

[0047] The present invention will now be described in detail with reference to the accompanying drawings.

[0048] This invention, based on Python, provides a genome-wide structural variation verification system, particularly suitable for structural variations in repetitive sequences. To further clarify the objectives, technical solutions, and advantages of this invention, the embodiments are described in detail below with reference to the accompanying drawings.

[0049] like Figure 1As shown, this invention provides a structural variation verification system suitable for genomic repetitive sequences, mainly comprising five modules: an input module, a sequence alignment module, a deduplication module, a structural variation evaluation module, and an output module. These modules are responsible for processing user input, performing hash-based repetition alignment of k-mer sequences, deduplicating the alignment results, evaluating structural variations, and outputting the evaluation results and visualizations. The following detailed descriptions of each module are intended to explain, not limit, the invention.

[0050] 1. Input module

[0051] The input module is used to input the sorted BAM file, the reference genome FASTA file, and the VCF file from the prior art, and sends the input files to the sequence alignment module;

[0052] The VCF file, or Variant Call Format, is a standard format for storing structural variation results; the FASTA file is a text-based format for identifying nucleic acid sequences, here referring to the reference genome sequence; and the BAM file is a binary file of the sequencing results.

[0053] The input module takes sorted BAM files, reference genome FASTA files, and VCF files output by various structural variation detection tools as input. It supports third-generation sequencing data from PacBio CLR (Continuous Long Reads), PacBio CCS (Circular Consensus Sequencing), and ONT (Oxford Nanopore Technologies), as well as gene alignment tools such as pbmm2, ngmlr, and minimap2. Furthermore, it can validate and evaluate structural variations based on results from mainstream structural variation detection tools such as CuteSV, Svim, Sniffles, and PBSV.

[0054] 2. Sequence alignment module

[0055] The sequence alignment module is used to perform sequence alignment on the received file using a kmer-based hash alignment algorithm. It obtains the starting coordinates of the matching sequence in the sequencing read and the reference genome sequence ref, the length of the matching sequence, and the orientation of the matching sequence. It then obtains the re-alignment results of the sequencing read containing structural variations, the reference genome sequence ref (the ref sequence is a fixed-length gene sequence), and the structural variation prediction sequence pre; and sends them to the deduplication module.

[0056] After inputting the reference genome's FASTA file, the sorted BAM file, and the VCF file output by the structural variation detection tool, information such as the chromosome, start and end coordinates, type, and length of each structural variation recorded in the VCF file is extracted. To more intuitively observe and analyze structural variations, an auxiliary length is extended to both sides of the structural variation's start and end coordinates, initially defaulting to 1000 bp. The actual auxiliary length is automatically adjusted based on the start and end coordinates of the sequencing reads covered within a given structural variation interval.

[0057] Based on the start and end coordinates of the structural variant, all m sequencing reads covered by a given structural variant s and the corresponding reference genome sequence ref can be retrieved from the BAM file. Furthermore, the structural variant s is inserted into the reference genome sequence ref to construct a predicted structural variant sequence pre. Figure 2 The predicted sequence pre is constructed using deletion, one of the gene structural variations, as an example.

[0058] Next, the sequencing reads were re-aligned with their corresponding reference genome sequences (ref, which are the original reference genome sequences) and structural variation prediction sequences (pre, which are the structural variation prediction sequences obtained after the recorded structural variations are inserted into the reference genome sequence). For the CCS and ONT sequencing data, a kmer sequence length of 31 was selected. The sequencing reads were traversed, and the starting coordinates of the kmer sequences were selected as their hash values ​​and stored in a hash table. Then, the inverse complementary sequences of the sequencing reads were traversed in the same way, and the kmer sequences and their corresponding starting coordinates were stored in a hash table. The hash table structure is shown in Table 1.

[0059] Table 1 Hash Table Structure

[0060] Key Value <![CDATA[read kmer1]]> <![CDATA[[kmer1 position1,kmer1 position2,...]]]> <![CDATA[read kmer2]]> <![CDATA[[kmer2 position1,kmer2 position2,...]]]> … … <![CDATA[read kmer n ]]> <![CDATA[[kmer n position1,Khmer n position2,...]]>

[0061] Next, the kmer sequence is traversed across the reference genome sequence (ref). For each new kmer sequence obtained, the hash table is checked for a matching kmer sequence. If a matching kmer sequence is found, the position of the matching kmer sequence in both the sequencing read and the reference genome sequence (ref) is moved 1 bp at a time to check if the next base is still the same, until the bases are different. The starting coordinates, length, and orientation of the matching sequence in both the sequencing read and the reference genome sequence (ref) are recorded (kmer sequences matching the sequencing read are in the positive direction, and kmer sequences matching the reverse complementary sequence of the sequencing read are in the negative direction). The result of this re-alignment is a series of matching start and end coordinates, represented as a line segment in a Cartesian coordinate system. The specific process of the hash re-alignment algorithm is as follows:

[0062] 1) Obtain the complementary sequencing sequence read_reverse_seq by reverse complementation of the sequencing sequence read_seq;

[0063] 2) For each kmer sequence read_kmer in the sequencing read_seq, add it to the hash table;

[0064] 3) For the complementary sequencing read_reverse_seq, perform the same processing as in step 2);

[0065] 4) Traverse all kmer sequences ref_kmer in the reference genome sequence ref_seq. If ref_kmer appears in the hash table in step 2), compare whether the next base of ref_kmer is the same as the next base of read_kmer in the hash table. If they are the same, expand the alignment sequence match_seq; otherwise, set match_seq = ref_kmer.

[0066] 3. Deduplication module

[0067] The deduplication module is used to remove duplicates from the sequence alignment results of the sequencing read containing structural variations output by the sequence alignment module, the reference genome ref sequence, and the structural variation prediction sequence pre, in order to eliminate the influence of repetitive sequences in the reference genome on structural variations; and send the deduplication results to the structural variation evaluation module.

[0068] Specifically, the human reference genome contains a large number of repetitive sequences, which severely affect the analysis of structural variants (SVs). Therefore, a novel approach was proposed: to eliminate the influence of repetitive sequences in the reference genome on regions of structural variation. In the sequence alignment module, the re-alignment results of the sequencing read containing structural variations with the reference genome's ref sequence and the predicted structural variation sequence pre were obtained. The deduplication results were then sent to the structural variation assessment module.

[0069] To obtain the coordinates of repetitive sequences within the reference genome sequence ref, the reference genome sequence ref is re-aligned with itself.

[0070] If the reference genome (ref) sequence does not contain repetitive sequences, the re-alignment result is a line segment of equal length to the reference genome (ref) sequence in a Cartesian coordinate system, and is represented as a line segment along the main diagonal in the sequence alignment segment plot. If the reference genome (ref) sequence contains repetitive sequences, the re-alignment result includes several segments of unequal length, which are represented as several line segments not along the main diagonal in the segment plot. The coordinate range of the refs of these repetitive sequences is recorded.

[0071] Next, the results of the re-alignment between the sequencing reads and the reference genome's ref sequences are traversed according to the ref coordinate order. In the segment plot (sequence alignment diagram), for segments to the left of structural variants, duplicates are removed while segments located on the main diagonal are retained; for segments to the right of SVs (structural variants), duplicates are removed while segments located on the line offset from the main diagonal by the SV length intercept are retained. Segments within the ref coordinate range of the repetitive sequences are removed, leaving segments containing only the structural variant regions. Traversing these segments in the segment plot, breakpoints appear between adjacent non-collinear segments.

[0072] A schematic diagram of the deduplication principle is shown below. Figure 3 As shown.

[0073] 4. Structural Variation Assessment Module

[0074] The structural variation evaluation module is used to evaluate the received deduplication results using a distance-based structural variation evaluation method, and then send the evaluation results to the output module.

[0075] Specifically, based on the results of the deduplication module, a novel distance-based structural variation assessment method is used to score and verify each structural variation.

[0076] First, segments generated from the alignment of sequencing reads with reference genome sequences (ref) and segments generated from the alignment of sequencing reads with predicted structural variation sequences (pre) are visualized to create segmentplots. The segments generated from the alignment are represented as several line segments in the segmentplot. If the two sequences are completely identical, the generated segmentplot will show only one line segment perfectly aligned with the main diagonal. Conversely, the line segments in the segmentplot will deviate from the main diagonal, and the magnitude of this deviation reflects the degree of difference between the two sequences. In the segmentplot corresponding to the sequencing read and reference genome sequence (ref), this difference indicates a structural variation in that region. In the segmentplot corresponding to the sequencing read and predicted structural variation sequence (pre), the magnitude of this difference indicates the degree of consistency between the structural variation present in that region and the structural variation described in the VCF file. This degree of difference is quantified by calculating the average distance between all line segments in the segmentplot and the main diagonal. The two degrees of difference are then normalized to achieve an assessment of the structural variation.

[0077] Suppose a structural variant *s* covers *m* sequencing reads, where the segment plot corresponding to the *i*th sequencing read has *n* line segments. Let *d* be the distance *d* of the *j*th line segment from the main diagonal. s,i,j for:

[0078] d s,i,j =1 / 3[(x s,i,j,start -y s,i,j,start )+(x s,i,j,mid -y s,i,j,mid )+(x s,i,j,end -y s,i,j,end )]

[0079] In the formula, x s,i,j,start y s,i,j,start Let x be the starting coordinate of the ref corresponding to the j-th line segment. s,i,j,mid y s,i,j,mid Let x be the coordinates of the midpoint of the j-th line segment, x is the midpoint of the read line segment. s,i,j,end y s,i,j,end Let ref be the coordinates of the endpoint of the j-th line segment, and let read be the coordinates of the endpoint of the line segment.

[0080] Define the average distance d in the segment plot corresponding to the structural variation s. s,i,j,avg for:

[0081]

[0082] Define the average distance between the sequencing read and the reference genome sequence ref as d. s,i,j,avg,ref , then d s,i,j,avg,ref ∈[0,+∞), a larger value indicates a structural variation in the region. The average distance d between the sequencing read and the structural variation prediction sequence pre is defined as... s,i,j,avg,predict , then d s,i,j,avg,ref ∈[0,+∞), the smaller the value, the more consistent the structural variation in the region is with the structural variation recorded in the VCF file. Normalizing both values ​​yields the score for this structural variation s. s,i :

[0083]

[0084] Furthermore, Score s,i ∈[0,1]. Score s,i A score of 0 indicates that there is no structural variation in the region; s,i A value of 1 indicates the presence of a structural variant within the region that is completely identical to the one recorded in the VCF file. In practical applications, sequencing errors or incomplete deduplication may result in noise points in certain segment plots, which will affect the score of the structural variant 's'. s,i The value of has some influence, therefore, the score of structural variation s is considered to be... s,i The closer the value is to 1, the more consistent the actual structural variation is with the VCF file record; the higher the score of structural variation s is. s,i The closer it is to 0, the less consistent it is with the actual situation and the structural variations recorded in the VCF file.

[0085] For a given structural variation s, calculate a score list of all m sequencing reads it covers. s,0 Score s,1 ,...,Score s,i ,...,Score s,m And the highest score among them:

[0086] Score s,highest =max([Score s,0 Score s,1 ,...,Score s,i ,...,Scores,m ]).

[0087] In addition, users can specify a scoring threshold for structural variations, which defaults to Score. threshold =0.8, assuming the score of the i-th sequencing read is... s,i >Score threshold If the sequence read is considered to be a supporting read for the structural variant s, then the sequence read is considered to be a supporting read for the structural variant s.

[0088] Therefore, the proportion of supporting reads for a given structural variation s s for:

[0089]

[0090] In the formula, Score threshold The user-defined score threshold for structural variation is a decimal between 0 and 1; when the score of structural variation s exceeds this threshold, the score is calculated. threshold At that time, the structural variation s is considered to be a qualified structural variation for evaluation;

[0091] Furthermore, the genotype of this structural variation s s for:

[0092]

[0093] 5. Output Module

[0094] The output module is used to output the received evaluation results;

[0095] The highest score in the structural variation assessment. s,highest The score of all sequencing reads covered by structural variations. s,0 Score s,1 ,...,Score s,i ,...,Score s,m The system supports the proportion of supported sequences for structural variations, genotyping, and writes all breakpoints within the structural variations into an additional column of the input VCF file. Furthermore, it outputs a segment plot visualizing each structural variation, with the default output being a deduplicated image, allowing users to more intuitively view the true situation of each structural variation. Users can specify whether to output images of the deduplication process.

[0096] To verify the correctness of the structural variation verification tool provided by this invention, experimental verification was conducted on both simulated and real samples.

[0097] 1. Simulated data

[0098] 21,000 structural variants were randomly inserted into the 23 human chromosomes, from chromosome 1 to chromosome X, including ten types of structural variants such as deletion, insertion, duplication, and inversion. Of these, 20,000 structural variants were used as positive samples and 1,000 as negative samples. Figure 4 The ROC curves for the above 21,000 sample data are shown.

[0099] The structural variation verification tool provided by this invention outputs a score for the structural variation, ranging from 0 to 1. The closer the score is to 1, the closer the structural variation is to the actual structural variation, meaning the verification result is more accurate. The recall rate of the simulated data changes with different user-defined score cutoffs (SV score cutoffs) as follows: Figure 5 As shown.

[0100] The structural variation verification tool provided by this invention is also applicable to genomic repetitive sequences. For verification, the 21,000 samples were screened to obtain 405 samples located in repetitive sequence regions. Figure 6 In Figures (a) and (b), the recall rate of structural variations in repeating regions changes with the scoring threshold. It can be seen that the structural variation verification tool provided by this invention has the same effect as structural variations in non-repeating regions when verifying structural variations within repeating regions.

[0101] 2. Real data

[0102] To compare the results of the structural variation verification tool (spotsv) provided by this invention with those of an existing tool (vapor), both tools were run on the HG002 structural variation dataset. Figure 7 The graph shows the recall rate of the two tools as a function of the scoring threshold in the HG002 sample set. It can be seen that, under different sequencing technologies and sequencing depths, the recall rate of spotsv, the structural variation verification tool provided in this invention, is significantly higher than that of vapor.

[0103] Figure 8 Images (a), (b), and (c) are visualizations of the structural variation tool provided by this invention, using deletion, one type of structural variation, as an example. Figure 1 The segment plot of the reference genome shows that there are a large number of repetitive sequences in this region. Figure 8Figure (b) shows the segment plot of the reference genome and sequencing reads before deduplication. Due to the presence of repetitive sequences, structural variations cannot be detected or identified. Figure 3 After deduplication, the segment plot of the reference genome and sequencing reads is free from the influence of repetitive sequences, allowing for a clear assessment of the true extent of structural variation. Figure 8 This demonstrates that the structural variation verification tool provided by this invention is applicable to repetitive sequences in the genome.

Claims

1. A structural variation verification system suitable for genomic repetitive sequences, characterized in that, include: The sequence alignment module is used to perform sequence alignment on the sorted BAM file, the reference genome's FASTA file, and the VCF file using a kmer-based hash alignment algorithm. It obtains the re-alignment results of the sequencing read containing structural variations with the reference genome's ref sequence and the structural variation prediction sequence pre, and sends them to the deduplication module. The deduplication module is used to deduplicatively compare the received sequencing read containing structural variations with the reference genome ref sequence and the structural variation prediction sequence pre, and then send the deduplication result to the structural variation evaluation module. The structural variation assessment module is used to evaluate the received deduplication results using a distance-based structural variation assessment method to verify the structural variations of genomic repetitive sequences. The fragments generated by the re-alignment of the sequencing read with the reference genome sequence ref and the fragments generated by the re-alignment of the sequencing read with the structural variant prediction sequence pre are visualized to generate sequence alignment maps. The fragments generated by the re-alignment are represented as several line segments in the sequence alignment map. If the sequencing read is completely identical to the reference genome sequence ref, there will be a line segment in the generated sequence alignment map that runs exactly along the main diagonal. Otherwise, the line segment in the sequence alignment map will deviate from the main diagonal, and the magnitude of the deviation reflects the degree of difference between the two sequences. By calculating the average distance between all line segments in the sequence alignment map and the main diagonal, the average distance in the sequence alignment map corresponding to the structural variant s, the average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the structural variant prediction sequence pre are obtained. The average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the structural variant prediction sequence pre are normalized to obtain the score of the structural variant s. The structural variant is evaluated based on the score of the structural variant s.

2. A method for verifying structural variations in genomic repetitive sequences, characterized in that, Includes the following steps: 1) For the sorted BAM file, the fasta file of the reference genome, and the VCF file, sequence alignment is performed using a kmer-based hash alignment algorithm to obtain the alignment results of the sequencing read containing structural variations with the ref sequence of the reference genome and the predicted sequence pre of structural variations. 2) Remove duplicates from the alignment results of sequencing reads containing structural variations with the reference genome ref sequences and the structural variation prediction sequences pre; 3) The deduplication results are evaluated using a distance-based structural variation assessment method to verify structural variations in genomic repetitive sequences. Specifically, the fragments generated from the re-alignment of the sequencing read with the reference genome sequence ref and the fragments generated from the re-alignment of the sequencing read with the predicted structural variation sequence pre are visualized to generate sequence alignment maps. The re-aligned fragments are represented as line segments in the sequence alignment map. If the sequencing read is completely identical to the reference genome sequence ref, then the generated sequence alignment map contains a line segment perfectly aligned with the main diagonal; otherwise, the line segment in the sequence alignment map deviates from the main diagonal, and the magnitude of the deviation reflects the degree of difference between the two sequences. By calculating the average distance between all line segments in the sequence alignment map and the main diagonal, the average distance in the sequence alignment map corresponding to the structural variation s, the average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the predicted structural variation sequence pre are obtained. The average distance between the sequencing read and the reference genome sequence ref, and the average distance between the sequencing read and the structural variant prediction sequence pre are normalized to obtain the score of the structural variant s. The structural variant is evaluated based on the score of the structural variant s.

3. The method for verifying structural variations in genomic repetitive sequences according to claim 2, characterized in that, The specific process of step 1) is as follows: Extract the start and end coordinates of each structural variation within the VCF file; Based on the start and end coordinates of the structural variant, all m sequencing reads covered by a certain structural variant s and the corresponding reference genome sequence ref are retrieved from the BAM file. The structural variant s is then inserted into the reference genome sequence ref to construct a structural variant prediction sequence pre. Then, the kmer sequence is traversed on the reference genome sequence (ref). For each new kmer sequence obtained, the hash table is indexed to see if there is a matching kmer sequence. If there is, the kmer sequence position that matches in the sequencing read and the reference genome sequence (ref) is moved by 1 bp to check if the next base is the same, until the bases of the two sequences are different. The starting coordinates of the matching sequence in the sequencing read and the reference genome sequence (ref), the length of the matching sequence, and the direction of the matching sequence are recorded. The re-alignment results of the sequencing read containing structural variations, the reference genome (ref) sequence, and the structural variation prediction sequence (pre) are obtained.

4. The method for verifying structural variations in genomic repetitive sequences according to claim 3, characterized in that, Hash tables are determined through the following process: The sequencing reads were re-aligned with the corresponding reference genome sequence ref and the structural variation prediction sequence pre, respectively. For the sequencing data of CCS and ONT, the length of the kmer sequence was selected as 31. The sequencing reads were traversed, and the starting coordinate of the kmer sequence was selected as the hash value of the kmer sequence and stored in a hash table. The reverse complementary sequence of the sequencing read is traversed, and the kmer sequence and its corresponding starting coordinates are stored in a hash table.

5. The method for verifying structural variations in genomic repetitive sequences according to claim 2, characterized in that, The specific process of step 2) is as follows: If there are no repetitive sequences in the reference genome ref sequence, the result of the re-alignment is a line segment with a length equal to the length of the reference genome ref sequence in a Cartesian coordinate system, and is represented as a line segment along the main diagonal in the sequence alignment diagram; If the reference genome ref sequence contains repetitive sequences, the re-alignment result includes several line segments of varying lengths, which are represented in the sequence alignment map as several line segments not along the main diagonal, recording the coordinate range of the ref of the repetitive sequence.

6. The method for verifying structural variations in genomic repetitive sequences according to claim 5, characterized in that, Step 2) also includes the following steps: The results of the re-alignment of the sequencing read with the reference genome's ref sequence are traversed according to the ref coordinate order of the repetitive sequences. In the sequence alignment diagram, for fragments located to the left of the structural variant, duplicates are removed while fragments located on the main diagonal are retained; for fragments located to the right of the structural variant, duplicates are removed while fragments located on the line that is offset from the main diagonal by the length intercept of the structural variant are retained.

7. The method for verifying structural variations in genomic repetitive sequences according to claim 2, characterized in that, The average distance in the sequence alignment plot corresponding to structural variant s Calculated using the following formula: Suppose a structural variant *s* covers *m* sequencing reads, where the sequence alignment graph corresponding to the *i*th sequencing read has *n* line segments, and the distance of the *j*th line segment from the main diagonal is... Calculated using the following formula: In the formula, , Let ref be the starting coordinate of the j-th line segment, and read be the starting coordinate of the line segment. , Let be the coordinates of the midpoint of the ref corresponding to the j-th line segment, and let be the coordinates of the midpoint of the read segment. , Let ref be the coordinates of the endpoint of the j-th line segment, and let read be the coordinates of the endpoint of the line segment.

8. The method for verifying structural variations in genomic repetitive sequences according to claim 2, characterized in that, The average distance between the sequencing read and the reference genome sequence (ref) and the average distance between the sequencing read and the structural variant prediction sequence (pre) are normalized to obtain the score of the structural variant s. : When the score of structural variation s is greater than the threshold At that time, the structural variation s is considered to be a qualified structural variation for evaluation.

Images