Method and system for detecting structural variations

Abstract

Methods and systems for detecting structural variations (SVs) from long read length sequencing data. The method comprises the following steps: encoding an SV signal with a compared read length from a binary comparison graph (BAM) file, wherein the SV signal is encoded in a matrix form; detecting one or more candidate regions from the encoded SV signal, wherein the one or more candidate regions comprise an SV signal above a predetermined signal level; clustering the compared read lengths in each of the one or more candidate regions to form one or more classes, respectively, in each of the one or more candidate regions; assembling the aligned reads within each class to generate one or more consensus sequences, respectively; and comparing the one or more consensus sequences with corresponding reference sequences to detect the SV.

Description

Technical Field

[0001] This invention relates broadly, but not exclusively, to methods and systems for detecting structural variations. Background Technology

[0002] Structural variations (SVs) are genomic variations longer than 50 base pairs (bps). SVs have been shown to have a significant impact on human diversity and are closely associated with genetic diseases. Although single nucleotide variants (SNVs) and small insertions / deletions (indels) are more common, SVs collectively alter a large portion of the human genome. In the era of short-read sequencing, existing research on SV recall has focused on methods such as read depth, inconsistent read pairing, split read alignment, and local assembly. These methods have facilitated large-scale genome studies, such as the 1000 Genomes Project. However, the limited read length of short-read sequencing hinders the detection of long SVs, and problematic alignments pose challenges, leading to potential false positives.

[0003] The advent of long-read sequencing technologies such as PacBio and Nanopore has made it possible to obtain longer reads spanning thousands of bps, revolutionizing SV detection. This advancement has significantly improved SV detection and facilitated more accurate analysis. Simple SVs can be classified as balanced (e.g., inversions and translocations) or unbalanced (e.g., deletions, insertions, and duplications), with unbalanced SVs altering the length of the genome. In addition to these simple SVs, complex SVs (also known as nested SVs) have been observed in real-world datasets and are gaining increasing attention in practical applications; they are particularly interesting. Complex SVs can be viewed as combinations of simple SVs in various orders and numbers. In particular, translocations can be considered a special type of complex SV based on deletions and insertions.

[0004] To address the challenges of SV detection, researchers have developed a variety of computational algorithms and tools. Most of these approaches focus on detecting simple SVs and can be broadly categorized into alignment-based methods and assembly-based methods.

[0005] Alignment-based methods typically involve three steps. First, SV signals are captured from individual reads using compact concise alignment information expressions (CIGAR) from binary alignment map (BAM) files. These signals are often referred to as SV features. SV features are represented as {latent type, start position, length}. Second, SV tags with the same latent type are clustered using thresholding or density-based methods. Within each cluster, SV tags are synthesized based on start position and length, thus identifying the detected latent SVs. Finally, a quality or confidence score is calculated for each latent SV using certain metrics such as the number of supporting reads. By applying a user-defined threshold to the calculated score, the SV is finally determined and reported in a variant call format (VCF) file.

[0006] Alignment-based methods are popular due to their simplicity and efficiency, especially in detecting simple SVs. However, problematic alignments can produce false positives and limit the possibility of detecting complex SVs because they determine the type of potential SV from the initial steps.

[0007] On the other hand, assembly-based methods overcome the limitations of alignment-based methods by assembling reads into longer shared sequences and aligning these shared sequences with a reference genome to detect SVs. By bypassing the alignment of individual reads, assembly-based methods are less affected by alignment results, but they are computationally expensive and unsuitable for large-scale data analysis. Summary of the Invention

[0008] Embodiments of the present invention provide a universal SV detector using third-generation sequencing data, comprising one or more of the following steps: encoding, detection, clustering, assembly, alignment, and final determination. Within each candidate region, reads within each cluster are assembled into a common sequence, which is mapped to a reference genome via a maximum exact match. The present invention combines the advantages of alignment-based and assembly-based methods and allows the detection of complex SVs by deferring the determination of SV types to the final alignment step. Specifically, these steps include: (a) extracting SV signals from aligned reads in a BAM file; (b) detecting regions with strong signal accumulation from the reads; (c) clustering corresponding reads within the candidate regions; (d) assembling reads within a cluster into a common sequence; (e) aligning the common sequence with a reference sequence; and (f) finally determining the SV based on the alignment results of step (d). The information in step (a) can be obtained through both internal and cross-alignment results. The detection step (b) applies a graph segmentation method to detect strong signals. The clustering step (c) is performed to address potential subtype issues. Assembly step (d) can be performed using the wtdbg2 assembly tool.

[0009] According to one aspect, this application provides a method for detecting structural variations (SVs) from long-read sequence data, comprising: encoding SV signals from aligned reads in a binary alignment map (BAM) file, wherein the SV signals are encoded in matrix form; detecting one or more candidate regions from the encoded SV signals, wherein the one or more candidate regions include SV signals above a predetermined signal level; clustering the aligned reads within each of the one or more candidate regions to form one or more classes accordingly; assembling the aligned reads within each of the one or more classes to generate one or more common sequences accordingly; and aligning the one or more common sequences with a corresponding reference sequence to detect SVs.

[0010] The step of aligning one or more common sequences with corresponding reference sequences to detect SVs may include using copmem2 to perform maximum exact match (MEM) alignment.

[0011] After aligning one or more common sequences with corresponding reference sequences to detect SVs, the method may further include filtering the results of the MEM alignment to determine SVs, wherein the filtering is based on inversions and / or gaps in one or more common sequences.

[0012] This method may include using the wtdbg2 assembly tool to assemble the aligned reads within each of one or more classes to generate one or more shared sequences accordingly.

[0013] This method may include using a graph cut method to detect one or more candidate regions from an coded SV signal.

[0014] On the other hand, this application provides a system for detecting structural variations (SVs) from long-read sequencing data, comprising: a processor module; and a memory module containing computer program code. The memory module and the computer program code can be configured, together with the processor module, to cause the system to perform at least the following operations: encoding SV signals from aligned reads in a binary alignment map (BAM) file, wherein the SV signals are encoded in matrix form; detecting one or more candidate regions from the encoded SV signals, wherein the one or more candidate regions include SV signals above a predetermined signal level; clustering the aligned reads within each of the one or more candidate regions to form one or more classes accordingly; assembling the aligned reads within each of the one or more classes to generate one or more common sequences accordingly; and aligning the one or more common sequences with a corresponding reference sequence to detect SVs.

[0015] The system can be further enabled to use copmem2 to perform maximum exact match (MEM) alignments to align one or more common sequences.

[0016] The system can be further configured to filter the results of the MEM alignment to determine the SV, wherein the filtering is based on inversions and / or gaps in one or more common sequences.

[0017] The system can be further enabled to use the wtdbg2 assembly tool to assemble the aligned reads within each of one or more classes to generate one or more common sequences accordingly.

[0018] The system can be further enabled to use graph cut methods to detect one or more candidate regions from the coded SV signal. Attached Figure Description

[0019] Embodiments of the invention will be better understood and readily apparent to those skilled in the art from the following written description, which is by way of example only and in conjunction with the accompanying drawings, in which:

[0020] Figure 1 This is a flowchart illustrating a method for detecting structural variations (SVs) from long-read sequencing data according to an embodiment.

[0021] Figure 2 An example of a complex SV including missing, duplicate, and inverted elements is shown.

[0022] Figure 3 This is a table illustrating the performance of the embodiment of the present invention in detecting SV compared to existing methods.

[0023] Figure 4 (a) is a screenshot of the Integrated Genome Viewer (IGV) that simulates DEL. Figure 4 (b) is a screenshot of the IGV simulating DUP.

[0024] Figure 5 The performance of candidate methods for detecting different types of complex SVs is shown by evaluating the region and type accordingly.

[0025] Figure 6 This is a flowchart illustrating a method for detecting structural variations (SVs) from long-read sequencing data according to an embodiment.

[0026] Figure 7 A schematic diagram of a computer system suitable for performing at least some steps of a method for detecting structural variations (SVs) from long-read sequencing data is shown. Detailed Implementation

[0027] Some of the parts described below are explicitly or implicitly presented with respect to the functional or symbolic representation of algorithms and operations in data stored in computer memory. These algorithmic descriptions and functional or symbolic representations are the means by which those skilled in the art of data processing most effectively communicate the substance of their work to others skilled in the art. An algorithm herein is generally considered to be a self-consistent sequence of steps that leads to a desired result. These steps are those that require physical manipulation of physical quantities such as electrical, magnetic, or optical signals that can be stored, transmitted, combined, compared, and otherwise manipulated.

[0028] It will be readily understood from the following text that, in this specification, discussions using terms such as “scan,” “calculate,” “determine,” “replace,” “generate,” “initialize,” “output,” and similar terms refer to the actions and processes of a computer system or similar electronic device that manipulate and transform data represented as physical quantities within the computer system into other data similarly represented as physical quantities within the computer system or other information storage, transmission, or display devices, unless otherwise specified.

[0029] This specification also discloses apparatus for performing these methods. Such apparatus may be specifically constructed for the desired purpose or may include a computer or other devices selectively activated or reconfigured by a computer program stored in the computer. The algorithms and displays presented herein are not inherently related to any particular computer or other apparatus. Various machines may be used with the programs taught herein. Alternatively, it may be appropriate to construct more specialized apparatus to perform the desired method steps. The structure of a conventional computer will become apparent from the description below.

[0030] Furthermore, this specification implicitly discloses a computer program, as it will be readily apparent to those skilled in the art that the various steps of the methods described herein can be implemented by computer code. The computer program is not intended to be limited to any particular programming language or its implementation. It will be understood that the teachings of the disclosure contained herein can be implemented using a variety of programming languages ​​and their encodings. Moreover, the computer program is not intended to be limited to any particular control flow. Many other variations of the computer program, using different control flows, exist without departing from the spirit or scope of the invention.

[0031] Furthermore, one or more steps of a computer program can be executed in parallel rather than sequentially. Such a computer program can be stored on any computer-readable medium. Computer-readable media can include storage devices suitable for interfacing with a computer, such as disks or optical discs, memory chips, or other storage devices. Computer-readable media can also include hardwired media, such as those exemplified in Internet systems, or wireless media, such as those exemplified in GSM, GPRS, 3G, or 4G mobile phone systems, as well as other wireless systems such as Bluetooth, ZigBee, and Wi-Fi. When loaded and executed on such a computer, the computer program effectively generates means for implementing the steps of the preferred method.

[0032] This invention can also be implemented as a hardware module. More specifically, in a hardware sense, a module is a functional hardware unit designed to be used in conjunction with other components or modules. For example, a module can be implemented using discrete electronic components, or it can be part of an overall electronic circuit such as an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA). Many other possibilities exist. Those skilled in the art will understand that the system can also be implemented as a combination of hardware and software modules.

[0033] Structural variations (SVs) have garnered significant attention in genomic research due to their importance in understanding genetic diseases, gene expression, and chromosome evolution. Advances in long-read sequencing technologies have enabled researchers to obtain even longer reads, increasing the potential for SV detection beyond single nucleotide variants. This specification relates to a universal pipeline for detecting SVs from long-read sequencing data. Specifically, the methods provided in this application encompass a combination of alignment-based and assembly-based approaches, enabling the detection of both simple SVs (insertions, deletions, inversions, and duplications) and potentially complex SVs formed by combinations of simple SVs.

[0034] Extensive simulation studies demonstrate that the embodiments of the present invention achieve superior performance compared to existing tools in SV detection. Notably, the embodiments of the present invention perform exceptionally well in complex SV detection. Furthermore, the embodiments of the present invention can be applied in conjunction with other tools to the HG002 dataset. The implementation of the present invention not only performs well in detecting benchmark insertions and missing values ​​but also identifies a variety of potentially complex SVs of interest.

[0035] This specification relates to a general-purpose SV detector using third-generation sequencing data. The pipeline includes one or more of the following six steps: encoding, detection, clustering, assembly, alignment, and final determination:

[0036] Encoding: This step involves extracting the SV signal from the aligned reads in the Binary Alignment Map (BAM) file using both internal and cross-alignment results.

[0037] Detection: This step involves using graph cut methods to detect regions with strong signal accumulation from the read length. For the sake of brevity, the details of the graph segmentation method will not be described in this specification. Further details can be found in the following reference: Zhou, X., Yang, C., Yu, W. "Moving object detection by detecting contiguous outliers in the low-rank representation", [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(3):597-610.

[0038] Clustering: In this step, within each candidate region, the corresponding read lengths are clustered to generate potential different subtypes.

[0039] Assembly: In this step, the read lengths within the class are assembled into a common sequence.

[0040] Comparison: In this step, the maximum exact match method is used to preserve as much variation information as possible.

[0041] Final determination: In this step, the SV call is finally determined by combining the above comparison results.

[0042] Embodiments of the present invention combine the advantages of alignment-based and assembly-based methods, and allow the detection of complex SVs by deferring the assignment of SV type markers to the final alignment step. To evaluate the performance of the embodiments, simulations were performed using the HG002 dataset, demonstrating the superiority of the present invention over prior art tools. For brevity, details regarding the HG002 dataset will not be described in this specification. Further details are available in: Liao, W., Asri, M., Ebler, J. et al., “Draft Reference for Human Pangenome”, [J]. Nature, 2023, 617: 312–324. Additionally, the embodiments can be applied to detect SVs in breast cancer data. By leveraging the advantages of alignment-based and assembly-based methods, the embodiments outperform existing tools in terms of recall, precision, and F1 score. The ability to identify complex SVs further enhances their practicality.

[0043] Each step of the production line plays a different role in the process, as will be discussed below. Figure 1 Please provide a detailed explanation. Figure 1This is a flowchart illustrating a method for detecting structural variations (SVs) from long-read sequencing data according to an embodiment.

[0044] 1. Encoding (Step 104):

[0045] To identify as many SVs as possible, embodiments of the present invention seek to retain as much information as possible in the early stages of detection. Unlike existing alignment-based SV detection tools, no predefined SV types are assigned at the outset. Instead, inconsistencies between read lengths and reference sequences are identified first. These inconsistencies encompass sequencing errors and potential variant tags.

[0046] A matrix-based approach was used to capture all inconsistencies. The reference sequence for each chromosome was encoded as a 4×reference sequence length matrix. The alignment result for each read (step 102) was also encoded as a 4×reference sequence length matrix. The start and end positions of each result were also recorded to determine their location within the reference. Differences between these matrices obtained through subtraction revealed all inconsistencies between the sample and the reference sequences.

[0047] The read length and specific matching results are stored accordingly in the sequence (SEQ) and CIGAR columns of the alignment file. For CIGAR information, the alignment result typically includes four options: "M" (match or no match), "I" (insertion), "D" (deletion), and "S" (soft cut). Soft cut indicates no match, and these bases are skipped. For example, as... Figure 1 The reference sequence shown is “TACTGAAGCAGATCC”, and the read length is “CTGAAGATC”. The CIGAR information captured from the alignment results is “5M3D4M”.

[0048] Different flags in the alignment results require different processing. Reads with FLAG="0" or "16" (indicating a single consecutive alignment result) are encoded. When the alignment result flag is "2048" (complementary comparison) or "2064" (2048+16, reverse complement + supplementary comparison), this indicates that different portions of the read were aligned to different positions in the reference. In this case, the read must also include alignment results with FLAG="0" or "16" (primary results). See reference below. Figure 1 FLAG = "0". Therefore, a single consecutive alignment result is encoded.

[0049] To reflect this positional relationship in the matrix, embodiments of the invention seek to encode the primary and supplementary results within the same matrix (step 104). If they have the same orientation, they are placed in the matrix based on their relative positional relationship, with the region between them represented by 0. If the primary and supplementary results have different orientations, the supplement is converted into their reverse complements to reflect this difference.

[0050] Compared to other existing SV callers, the matrix-based encoding method according to embodiments of the present invention preserves all inconsistencies, thereby increasing the possibility of detecting more and more complex SVs. Let S j ={0,1} indicates whether the j-th position corresponding to the reference is a mutation, r ij =(a ij ,t ij ,c ij g ij ) indicates the encoding result at position j in the i-th read length, and ref j Indicates the encoding result at position j in the reference.

[0051] 2. Detection (Step 106):

[0052] After encoding the signal in matrix form (step 104), embodiments of the present invention seek to detect candidate regions with strong SV signals. Using the notation defined above, the problem is formulated as follows:

[0053]

[0054] The above formula consists of three parts. The first part represents when a position is labeled as non-mutant (S j The loss is calculated when the L2 norm is 0, and the L2 norm is used to quantify the inconsistency between the read length and the reference sequence. The second part is labeled as a variant (S) at a position. j The first part is the penalty term when β = 1, where the parameter β acts as a threshold to support the evidence. The model weighs the first two terms against the label 1 and 0. The third part is a penalty term used to penalize different adjacent labels. The aim is to further smooth the label sequence and facilitate the detection of continuous variations. This formula enables the identification of candidate regions with strong signals for further analysis.

[0055] 3. Clustering:

[0056] In each candidate region, there may be multiple reads representing different SVs or genotypes belonging to different subtypes within that region. If the user chooses to utilize prior information obtained during the encoding step to aid detection, candidate regions can be categorized into two types: regions with easily identifiable features and regions with complex, easily confused features (typically containing complex SVs and some simple SVs). See step 108, where the workflow differs for regions with different labels (step 109) and regions with complex, easily confused labels (step 110).

[0057] For regions with easily identifiable features, frequency-based clustering is used to determine the subtypes or genotypes of SVs within them. Specifically, embodiments of the invention cluster tags within a region based on start position, length, and type. Tags with the same type and similar length and position are grouped together. By analyzing the number of supporting reads, embodiments of the invention can determine their genotypes and then perform a final call on their corresponding SVs. This is illustrated at step 109.

[0058] For regions with complex and mixed characteristics, alignment results are often too chaotic to discern what is happening in these regions. Therefore, embodiments of the present invention use assembly and alignment techniques to determine sequence structure. However, directly assembling reads corresponding to different subtypes can result in problematic shared sequences that fail to reflect any subtype characteristics.

[0059] To address this issue, a clustering step is introduced in each region (step 110). Given that SVs typically cover many bases, if multiple subtypes exist within an SV, reads corresponding to different subtypes will exhibit significant bias within that region. However, if there is only one subtype, the differences between reads are primarily due to other factors such as sequencing errors or mapping errors. Importantly, embodiments of the invention allow only one subtype per region, which is the typical case. This task differs slightly from the classical clustering problem but remains within the framework of statistical hypothesis testing. The null hypothesis is that a pair of reads originates from a single subtype, while the alternative hypothesis is that a pair of reads does not originate from a single subtype.

[0060] To construct the test statistics, the difference between each pair of reads is directly calculated. These differences are then compared to a user-defined threshold for filtering. The testing process is performed in order of read length from longest to shortest, ensuring that each read is typically checked only once. Therefore, this algorithm runs efficiently with a time complexity of O(n), where n is the number of reads in the candidate region.

[0061] Figure 2An example of a simulated complex SV including missing, duplicate, and inverted parts is shown, where part A is first repeated, then the repeated parts A and B are inverted, and finally part C is deleted. This complex SV occurs alongside different types of simple SVs and is difficult for existing tools to detect accurately. Embodiments of the present invention can provide an accurate invocation, as shown at the bottom of the figure.

[0062] 4. Assembly (Step 112):

[0063] For each class of reads, the assembly tool wtdbg2 is used to generate a common sequence. For brevity, the details of the assembly tool wtdbg2 will not be described in this specification. Further details can be found in the following reference: Ruan, J., Li, H., “Fast and accurate long-read assembly with wtdbg2”, [J]. Natural Methods, 2020, 17: 155–158. This assembly step allows the user to obtain a longer, continuous sequence for representing the combined information from the reads within each class. The wtdbg2 tool is also used in the Debreak tool to detect long insertions by assembling reads around them (Chen, Y., Wang, AY, Barkley, CA et al., “Deciphering the exact breakpoints of structural variations using long sequencing reads with debreak”, [J]. Nature Communications, 2023, 14(1): 283.).

[0064] 5. Comparison (Step 114):

[0065] For the assembled common sequence, it is compared with a reference genome to identify variations present within the class. In embodiments of the invention, typically only one common sequence is used for each class that potentially contains one or more SVs.

[0066] The Maximum Exact Match (MEM) method is used for the alignment step. For brevity, the details of MEM will not be described further in this specification. Further details can be found in the following literature: Grabowski, S., Bieniecki, W., “copMEM2: robust and scalable maximum exact match finding”, [J]. Bioinformatics, 2023, 39(5): btad313. MEM identifies and reports the maximum exact match between the shared sequence and the reference sequence, while ignoring some gaps. This method, despite producing some gaps, preserves as much information as possible and allows users to more accurately infer the differences between the shared sequence and the reference sequence. In addition, if there are potential complex SVs, MEM also enables users to capture information about potential complex SVs. According to the implementation, the copmem2 tool is used for MEM alignment. For brevity, the details of the copmem2 tool will not be described in this specification. Further details can be found in the following literature: Grabowski, S., Bieniecki, W., “copMEM2: robust and scalable maximum exact match finding”, [J]. Bioinformatics, 2023, 39(5): btad313.

[0067] 6. Final determination:

[0068] The alignment results obtained using MEM are transformed into the final SV call (see...). Figure 1 (The “SV call set” in the text). This step is similar to the encoding step, but with additional consideration for inversions and gaps. When using MEM, both forward matches and inversion matches are reported, along with their positions in the shared and reference sequences and the match length. However, gaps may exist in the alignment in addition to SVs due to various factors such as sequencing errors, mapping errors, single nucleotide polymorphisms (SNPs), and small indels. These gaps are typically short, covering only a few base pairs (bps), but they can introduce gaps in both the shared and reference sequences. The difference in gap length between the shared and reference sequences is generally less than 50 base pairs. In embodiments of the invention, gaps not introduced by SVs are filled, and inversions are converted to forward sequences. Subsequently, the final SV is determined based on the aforementioned encoding steps.

[0069] The final SV may not belong to common SV types, such as missing and insertion. Complex SVs involving combinations of ordinary SVs can be directly inferred from MEM results. For potentially complex SVs, the order of combinations of simple SVs is very important, which can also be inferred from MEM.

[0070] In summary, embodiments of the present invention relate to a universal pipeline for detecting SVs from long-read sequencing data, and encompass a combination of alignment-based and assembly-based methods, enabling the detection of both simple SVs (insertions, deletions, inversions, and duplications) and complex SVs formed by combinations of simple SVs. Using both internal and cross-alignment results, SV signals are extracted from aligned reads in a BAM file for encoding. Next, regions with strong signal accumulation are detected from the reads using a graph segmentation method. Within each candidate region, the corresponding reads are clustered to account for potential different subtypes. The reads within each cluster are then assembled into a common sequence, which is mapped to a reference genome via maximum exact matching.

[0071] To evaluate the performance of the embodiments of the present invention, simulations were performed using the HG002 dataset, demonstrating its superiority over existing tools. Furthermore, the embodiments can be applied to detect SVs in breast cancer data. Figure 3 This is a table illustrating the performance of implementations of the present invention in detecting simple SVs. By utilizing the advantages of both alignment-based and assembly-based methods, the embodiments (see the row marked "this application") outperform existing tools (such as SVIM, Sniffles, PBSV, CuteSV, Debreak) in terms of recall, precision, and F1 score. The ability to identify complex SVs further enhances its practicality.

[0072] Figure 4 (a) is a screenshot of the Integrated Genome Viewer (IGV) that simulates DEL, detected only by an embodiment of the present invention. Figure 4 (b) is a screenshot of IGV for simulated DUP detected only by embodiments of the present invention. It should be noted that the content in parentheses represents its identifier in the Genome Variation Database (DGV).

[0073] Figure 5 The performance of candidate methods for detecting different types of complex SVs is shown by evaluating the region and type accordingly.

[0074] Figure 6This is a flowchart illustrating a method for detecting structural variations (SVs) from long-read sequencing data. The method includes step 602, which encodes SV signals from aligned reads in a binary alignment map (BAM) file, wherein the SV signals are encoded in matrix form. Step 604 involves detecting one or more candidate regions from the encoded SV signals, wherein the one or more candidate regions include SV signals above a predetermined signal level.

[0075] Step 606 involves clustering the aligned reads within each of the one or more candidate regions to form one or more classes accordingly. There can be more than one candidate region. As an example, suppose there are two candidate regions: A and B. (Reference) Figure 1 The figure is labeled "A". It should be noted that, for the sake of brevity, Figure 1 "B" is not shown. There may be multiple aligned reads in each candidate region. For example, region A has five aligned reads. The five aligned reads in region A are clustered together to form classes X1 and X2. See also Figure 1 The "X1" and "X2" in the text. Different types of subtypes, also known as haplotypes, may exist within the same region. For any given region, reads are clustered so that they can correctly correspond to their own haplotypes.

[0076] Step 608 involves assembling the aligned reads within each of one or more classes to generate one or more shared sequences accordingly. In class X1, three aligned reads are assembled to form a shared sequence C1. For brevity, Figure 1 The two compared reads in X2 are not shown in the figure.

[0077] Step 610 involves aligning one or more shared sequences with corresponding reference sequences to detect SVs. The shared sequence C1 is aligned with the reference sequence R1. Generally, there is only one shared sequence. In the case of multiple shared sequences, each shared sequence can have its own corresponding reference sequence. Finally, SVs can be detected by comparing C1 and R1.

[0078] The step of aligning one or more common sequences with corresponding reference sequences to detect SVs may include using copmem2 to perform maximum exact match (MEM) alignment.

[0079] After aligning one or more shared sequences with corresponding reference sequences to detect SVs, the method further includes filtering the results of the MEM alignment to determine the SVs. The filtering is based on inversions and / or gaps in the one or more shared sequences.

[0080] The wtdbg2 assembly tool can be used to assemble aligned reads within each of one or more classes to generate one or more common sequences accordingly.

[0081] One or more candidate regions can be detected from an coded SV signal using graph segmentation methods.

[0082] According to embodiments of the present invention, a method for detecting structural variants (SVs) from long-read sequencing data is provided, comprising one or more of the following steps: (a) encoding SV signals from aligned reads in a BAM file to capture inconsistencies between the reads and a reference genome; (b) detecting candidate regions with strong SV signals by minimizing loss at non-variant sites and penalizing variant sites; (c) clustering reads within each candidate region to determine SV subtypes or genotypes; (d) assembling reads within each cluster into a common sequence using an assembly tool; (e) aligning the assembled common sequence to a reference genome using a maximum exact match method to identify variants present within the cluster; and (f) finalizing SV calls by transforming the alignment results into final SV calls, taking into account inversions and gaps, and determining the final SV type based on the encoding steps.

[0083] The encoding step may involve encoding the SV signal in matrix form to represent inconsistencies between the read length and the reference genome, and obtaining all inconsistencies between the sample and the reference sequence by subtracting the two matrices.

[0084] The detection steps may involve a formulaic mathematical optimization problem to minimize the loss at non-mutated locations, penalize mutated locations, and penalize distinct neighboring labels in order to detect candidate regions with strong SV signals.

[0085] The alignment step may involve using a maximum exact match method to align the assembled common sequence with a reference genome, thereby capturing the maximum exact match and ignoring some gaps to accurately infer the differences between the common sequence and the reference genome.

[0086] The final determination step may involve transforming the alignment results obtained using the maximum exact match into a final SV call.

[0087] According to embodiments of the present invention, a system for detecting structural variations (SVs) from long-read sequencing data is provided, comprising: a processor module; and a memory module containing computer program code. The memory module and the computer program code are configured, together with the processor module, to cause the system to perform at least the following operations: encoding SV signals from aligned reads in a binary alignment map (BAM) file, wherein the SV signals are encoded in matrix form; detecting one or more candidate regions from the encoded SV signals, wherein the one or more candidate regions include SV signals above a predetermined signal level; clustering the aligned reads within each of the one or more candidate regions to form one or more classes accordingly; assembling the aligned reads within each of the one or more classes to generate one or more common sequences accordingly; and aligning the one or more common sequences with a corresponding reference sequence to detect SVs.

[0088] The system can be further configured to use copmem2 to perform maximum exact match (MEM) alignments to compare one or more common sequences.

[0089] The system can be further configured to filter the results of MEM alignments to determine SVs, where the filtering is based on inversions and / or gaps in one or more common sequences.

[0090] The system can be further enabled to use the wtdbg2 assembly tool to assemble the aligned reads within each of one or more classes to generate one or more common sequences accordingly.

[0091] The system can be further enabled to use graph segmentation methods to detect one or more candidate regions from the coded SV signal.

[0092] This system can be further configured to perform / fulfill other methods / steps described herein.

[0093] According to an embodiment, a computer-implemented method for detecting structural variations (SVs) from long-read sequencing data is provided, the method comprising the following steps.

[0094] Figure 7 A schematic diagram of a computer system suitable for performing at least some steps of a method for detecting structural variations (SVs) from long-read sequencing data is shown.

[0095] Figure 7 An exemplary computing device 700 is depicted, which is hereinafter interchangeably referred to as computer system 700, wherein one or more such computing devices 700 can perform / enforce methods for detecting structural variations and / or other methods / steps described herein. The following description of the computing device 700 is provided by way of example only and is not intended to be limiting.

[0096] like Figure 7 The example computing device 700 shown includes a processor 704 for executing software routines. Although a single processor is shown for clarity, the computing device 700 may also include a multiprocessor system. The processor 704 is connected to a communication infrastructure 706 to communicate with other components of the computing device 700. The communication infrastructure 706 may include, for example, a communication bus, a crossbar switch, or a network.

[0097] The computing device 700 also includes a main memory 708 such as random access memory (RAM) and an auxiliary memory 710. The auxiliary memory 710 may include, for example, a hard disk drive 712 and / or a removable storage drive 714, which may include a floppy disk drive, magnetic tape drive, optical disk drive, or similar device. The removable storage drive 714 reads from and / or writes to the removable storage unit 718 in a well-known manner. The removable storage unit 718 may include a floppy disk, magnetic tape, optical disk, or similar device read from and written to by the removable storage drive 714. Those skilled in the art will understand that the removable storage unit 718 includes a computer-readable storage medium in which computer-executable program code instructions and / or data are stored.

[0098] In an alternative embodiment, the auxiliary memory 710 may additionally or alternatively include other similar means for allowing computer programs or other instructions to be loaded into the computing device 700. Such means may include, for example, removable storage unit 722 and interface 720. Examples of removable storage unit 722 and interface 720 include program cassettes and cassette interfaces (such as those found in video game console devices) that allow software and data to be transferred from removable storage unit 722 to computer system 700, removable storage chips (such as EPROM or PROM) and associated sockets, and other removable storage units 722 and interfaces 720.

[0099] The computing device 700 also includes at least one communication interface 724. The communication interface 724 allows software and data to be transferred between the computing device 700 and external devices via a communication path 726. In various embodiments of the invention, the communication interface 724 permits data to be transferred between the computing device 700 and a data communication network, such as a public or private data communication network. The communication interface 724 can be used to exchange data between different computing devices 700 forming part of an interconnected computer network. Examples of the communication interface 724 may include a modem, a network interface (such as an Ethernet card), a communication port, an antenna with associated circuitry, and similar devices. The communication interface 724 may be wired or wireless. The software and data transmitted via the communication interface 724 are in the form of signals, which may be electronic, electromagnetic, optical, or other signals that can be received by the communication interface 724. These signals are provided to the communication interface via the communication path 726.

[0100] like Figure 7 As shown, the computing device 700 also includes a display interface 702 and an audio interface 732. The display interface 702 performs operations for presenting images to an associated display 730, and the audio interface 732 performs operations for playing audio content via an associated speaker 734.

[0101] As used herein, the term "computer program product" may refer in part to removable storage unit 718, removable storage unit 722, hard disk installed in hard disk drive 712, or a carrier wave carrying software via communication path 726 (wireless link or cable) to communication interface 724. A computer-readable storage medium is any non-transitory tangible storage medium that provides recorded instructions and / or data to computing device 700 for execution and / or processing. Examples of such storage media include floppy disks, magnetic tapes, CD-ROMs, DVDs, and Blu-ray discs. TM Disks, hard disk drives, ROMs or integrated circuits, USB storage devices, magneto-optical disks, or computer-readable cards such as PCMCIA cards, and similar devices, whether such devices are internal or external to the computing device 700. Examples of temporary or intangible computer-readable transmission media that may also participate in providing software, applications, instructions, and / or data to the computing device 700 include radio or infrared transmission channels and network connections to another computer or networked device, as well as the Internet or intranets containing email transmissions and information recorded on websites, and similar devices.

[0102] A computer program (also referred to as computer program code) is stored in main memory 708 and / or auxiliary memory 710. The computer program may also be received via communication interface 724. When executed, this computer program enables computing device 700 to perform one or more features of the embodiments discussed herein. In various embodiments, when executed, the computer program enables processor 704 to perform various features of the embodiments described above. Therefore, this computer program represents a controller of computer system 700.

[0103] The software may be stored in a computer program product and loaded into a computing device 700 using a removable storage drive 714, a hard disk drive 712, or an interface 720. Alternatively, the computer program product may be downloaded to the computer system 700 via a communication path 726. When executed by a processor 704, the software causes the computing device 700 to perform the functions of the embodiments described herein.

[0104] It should be understood that, Figure 7 The embodiments described are presented by way of example only. Therefore, in some embodiments, one or more features of the computing device 700 may be omitted. Furthermore, in some embodiments, one or more features of the computing device 700 may be combined together. Additionally, in some embodiments, one or more features of the computing device 700 may be separated into one or more components.

[0105] Those skilled in the art will understand that various changes and / or modifications can be made to the invention illustrated in particular embodiments without departing from the spirit or scope of the invention as broadly described. Therefore, the present embodiments are to be considered illustrative rather than restrictive in all respects.

Claims

1. A method for detecting structural variations (SVs) from long-read sequencing data, comprising: The SV signal of the compared read length from the binary alignment map (BAM) file is encoded, wherein the SV signal is encoded in matrix form; Detect one or more candidate regions from the encoded SV signal, wherein the one or more candidate regions include SV signals above a predetermined signal level; The compared reads within each of the one or more candidate regions are clustered to form one or more classes accordingly; The aligned reads within each of the one or more classes are assembled to generate one or more common sequences accordingly; The one or more shared sequences are compared with the corresponding reference sequences to detect SV.

2. The method of claim 1, wherein the step of comparing the one or more common sequences with corresponding reference sequences to detect SVs comprises: Use copmem2 to perform maximum exact match (MEM) alignment.

3. The method of claim 2, wherein after comparing the one or more common sequences with corresponding reference sequences to detect SV, the method further comprises: The results of the MEM alignment are filtered to determine the SV, wherein the filtering is based on inversions and / or gaps in the one or more common sequences.

4. The method of claim 1, further comprising using the wtdbg2 assembly tool to assemble the aligned reads in each of the one or more classes to generate the one or more common sequences accordingly.

5. The method of claim 1, further comprising using a graph cut method for detecting the one or more candidate regions from the coded SV signal.

6. A system for detecting structural variations (SVs) from long-read sequencing data, comprising: Processor module; as well as A memory module containing computer program code; The memory module and the computer program code are configured, together with the processor module, to cause the system to perform at least the following operations: The SV signal of the compared read length from the binary alignment map (BAM) file is encoded, wherein the SV signal is encoded in matrix form; Detect one or more candidate regions from the encoded SV signal, wherein the one or more candidate regions include SV signals above a predetermined signal level; The compared reads within each of the one or more candidate regions are clustered to form one or more classes accordingly; The aligned reads within each of the one or more classes are assembled to generate one or more common sequences accordingly; The one or more shared sequences are compared with the corresponding reference sequences to detect SV.

7. The system of claim 6, further comprising using copmem2 to perform maximum exact match (MEM) alignment for aligning the one or more common sequences.

8. The system of claim 7, further comprising filtering the results of the MEM alignment to determine the SV, wherein the filtering is based on inversions and / or gaps in the one or more common sequences.

9. The system of claim 6, further comprising using the wtdbg2 assembly tool to assemble the aligned reads within each of the one or more classes to generate the one or more common sequences accordingly.

10. The system of claim 6, further comprising using a graph cut method for detecting the one or more candidate regions from the coded SV signal.

Images