Web-based visualization analysis method and system for tumor gene mutation detection by whole exome sequencing

Abstract

The application relates to the technical field of gene sequencing data processing and bioinformation analysis, in particular to a Web-based whole-exome sequencing tumor gene mutation detection visual analysis method and system. The system collects user sequencing data and a reference genome version through a Web interactive interface; a program is called to perform quality control cleaning and evaluation on the data, and a visual report is generated; sequence alignment is completed based on the reference genome, and a variation site is identified through algorithm iteration; biological annotation of the variation is combined with a database, a candidate pathogenic mutation set is screened out in multiple levels according to a strategy, the candidate set is projected to a visual interface, a site state is confirmed or removed in response to a manual checking instruction, and a final gene mutation detection report is generated. The application greatly simplifies the whole-exome sequencing data processing procedure, makes it easy for clinical doctors or researchers without bioinformation background to start, and improves the popularization rate and work efficiency of tumor gene detection work.

Description

Technical Field

[0001] This invention relates to the field of gene sequencing data processing and bioinformatics analysis technology, specifically to a Web-based method and system for visual analysis of tumor gene mutation detection using whole-exome sequencing. Background Technology

[0002] With the rapid development of high-throughput sequencing technology, whole-exome sequencing has been widely used in tumor gene mutation detection; the massive amount of sequencing data and the complexity of bioinformatics analysis have increased significantly, which puts forward higher requirements for data processing and result interpretation.

[0003] Currently, gene mutation detection usually relies on complex command-line tools or automated processes, which have a high technical threshold and are often difficult for clinicians or researchers to operate directly. Although existing automated analysis processes can handle large-scale data, most of them lack intuitive visualization and interaction, and relying solely on algorithm models cannot completely eliminate interference such as sequencing artifacts, strand bias, and end effects. This black-box analysis mode often results in false positive sites in the test results, and there is a lack of convenient manual verification methods, making it difficult to meet the stringent accuracy requirements of clinical applications.

[0004] Therefore, how to lower the threshold of bioinformatics analysis and effectively eliminate false positive interference in algorithm interpretation to improve the accuracy of detection results has become an urgent problem to be solved in this field.

[0005] The information disclosed in the background section above is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention discloses a web-based method and system for visual analysis of tumor gene mutation detection using whole-exome sequencing. Specifically, the technical solution of this invention is as follows:

[0007] Web-based visualization and analysis methods for whole-exome sequencing tumor gene mutation detection include:

[0008] The system collects raw sequencing data uploaded by users and information on the reference genome version selected by the users through a web interface.

[0009] The preset quality control program is invoked to perform quality assessment and cleaning on the raw sequencing data, generating quality-controlled sequencing data and a visual quality control report;

[0010] Based on the reference genome version information, the quality control sequencing data is mapped to the corresponding reference genome using an alignment algorithm to generate a sequence alignment file.

[0011] The mutation detection algorithm is invoked to scan the sequence alignment file, identify genomic mutation sites, and generate an original mutation record file containing mutation information.

[0012] Biological information annotation is performed on the mutation sites in the original mutation record file using a preset mutation annotation program and database, and the annotated mutation sites are screened at multiple levels according to a preset filtering strategy to generate a candidate pathogenic mutation set.

[0013] The candidate pathogenic mutation set is projected onto the visual verification interface. In response to the received manual inspection and verification command, the site status in the candidate pathogenic mutation set is confirmed or removed, and the final gene mutation detection report is generated.

[0014] Preferably, the step of calling a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generating quality-controlled sequencing data and a visualized quality control report includes:

[0015] The raw sequencing data were scanned in a single operation using the Fastp quality control algorithm.

[0016] Calculate the average read length, repetition rate, peak value of inserted fragments, total number of bases, Q20 base ratio, Q30 base ratio and GC content of the raw sequencing data to generate raw data quality control indicators;

[0017] Perform low-quality read filtering and adapter sequence removal operations to generate the quality control sequencing data;

[0018] Based on the post-quality control sequencing data, the post-quality control data indicators are calculated, and the original data quality control indicators and the post-quality control data indicators are rendered into the visual quality control report in HTML and JavaScript object representation JSON format.

[0019] Preferably, the step of mapping the quality control sequencing data to the corresponding reference genome using an alignment algorithm based on the reference genome version information to generate a sequence alignment file includes:

[0020] In response to the reference genome version information, the corresponding reference genome index file is loaded;

[0021] The BWA alignment tool was invoked to align the quality-controlled sequencing data with the loaded reference genome, generating initial alignment results.

[0022] The initial alignment results are format-converted and sorted to generate the sequence alignment file in binary alignment format (BAM).

[0023] Preferably, the step of calling the variant detection algorithm to perform a traversal scan of the sequence alignment file, identify genomic variant sites, and generate an original variant record file containing variant information includes:

[0024] Determine the variant detection mode, which includes a single-sample mode or a paired-sample mode;

[0025] The Vardict mutation detection algorithm is invoked to read the sequence alignment file;

[0026] Insertion / deletion variants, single nucleotide variants, and complex variants are identified in the sequence alignment file, and local re-alignment is performed to correct allele frequencies.

[0027] The identified variant sites and their corresponding genotype information and sequencing depth information are written into the original variant record file of the VCF variant calling format.

[0028] Preferably, the step of annotating the variant sites in the original variant record file with biological information using a preset variant annotation program and database includes:

[0029] The mutation annotation program is invoked to parse the original mutation record file;

[0030] The variant sites in the original variant record file are matched with preset frequency databases and functional databases; wherein, the frequency databases include the 1000 Genomes Database and the Exome Aggregation Consortium Database, and the functional databases include the COSMIC Tumor Mutation Database and the ClinVar Clinical Variance Database.

[0031] Add population frequency information, functional prediction information, and pathogenicity classification information to the successfully matched variant sites to generate an annotated variant list.

[0032] Preferably, the step of performing multi-level screening of annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set includes:

[0033] First-level filtering: Traverse the variant list after the annotation, remove variant sites with frequencies higher than the preset background frequency threshold in the frequency database, and retain rare variant sites;

[0034] Second-level screening: Examine the genomic regions and functional impacts of the rare variant sites, retain sites located in exon regions or splice regions and whose mutation type is non-synonymous mutation, and generate a set of functionally relevant variants;

[0035] Third-level screening: Obtain the sequencing support number and variant allele frequency for each site in the functionally relevant variant set, and perform the following classification and screening logic:

[0036] When the mutation type is determined to be germline mutation, the site with a sequencing support number greater than a preset first depth threshold and a mutation allele frequency greater than a preset first frequency threshold is retained, and the site is included in the candidate pathogenic mutation set.

[0037] When the mutation type is determined to be a somatic mutation, the site with a sequencing support number greater than the preset second depth threshold and the mutation allele frequency greater than the preset second frequency threshold is retained, and the site is included in the candidate pathogenic mutation set.

[0038] Sites that fail to meet the above sequencing support number or variant allele frequency conditions are filtered out.

[0039] Preferably, the step of projecting the candidate pathogenic mutation set onto the visual verification interface and, in response to a received manual inspection and verification instruction, confirming or removing the site status in the candidate pathogenic mutation set includes:

[0040] Call the Genome Browser IGV interface to load the sequence alignment file and the candidate pathogenic mutation set in the visualization verification interface;

[0041] The alignment of reads at the mutation sites is displayed, including alignment quality values, positive and negative strand distribution, and read end position distribution.

[0042] In response to a manual determination that the variant site has chain bias or is located in a high mismatch region at the end of a read, a removal instruction is received, the variant site is marked as a false positive and removed from the results;

[0043] In response to manual determination that the alignment quality of the variant site is qualified and there are no obvious sequencing artifacts, a confirmation instruction is received, and the variant site is marked as the final pathogenic mutation.

[0044] A web-based whole-exome sequencing tumor gene mutation detection visualization and analysis system, comprising:

[0045] The data acquisition module is used to collect raw sequencing data uploaded by users and the reference genome version information selected by users through a web interactive interface;

[0046] The quality control module is configured to call a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generate quality-controlled sequencing data and a visual quality control report.

[0047] The sequence alignment module is configured to map the quality control sequencing data to the corresponding reference genome based on the reference genome version information using an alignment algorithm, and generate a sequence alignment file.

[0048] The variant detection module is configured to call a variant detection algorithm to perform a traversal scan of the sequence alignment file, identify genomic variant sites, and generate an original variant record file containing variant information.

[0049] The annotation and filtering module is configured to use a preset variant annotation program and database to annotate the variant sites in the original variant record file with biological information, and to perform multi-level filtering on the annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set.

[0050] The visualization verification module is configured to project the candidate pathogenic mutation set onto the visualization verification interface, and in response to the received manual inspection and verification command, to confirm or remove the site status in the candidate pathogenic mutation set and generate the final gene mutation detection report.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. This invention constructs a web-based interactive analysis environment, realizing the entire process from data upload and parameter selection to report generation through a graphical interface; it uses the Fastp algorithm to generate visual quality control reports in HTML and JSON formats, intuitively displaying key indicators such as Q20 / Q30 and GC content, enabling users to complete professional data quality control and analysis without needing to master Linux command lines or complex bioinformatics programming knowledge; this design greatly simplifies the processing flow of whole exome sequencing data, making it easy for clinicians or researchers without a bioinformatics background to get started, improving the popularization rate and efficiency of tumor gene detection work;

[0053] 2. This invention implements a rigorous three-level filtering strategy. By integrating frequency databases such as the 1000 Genomes Database and ExAC, as well as clinical functional databases such as COSMIC and ClinVar, it achieves in-depth annotation and cleaning of variant sites. The system first removes high-frequency background noise, identifies non-synonymous mutations in exons and splice regions, and further sets differentiated sequencing depth and allele frequency thresholds according to the different characteristics of germline and somatic cell variations. This multi-dimensional hierarchical screening mechanism can accurately remove invalid variants from massive amounts of data, ensuring that the final output set of candidate pathogenic mutations has extremely high biological significance and clinical reference value.

[0054] 3. To address the issue of false positives that are difficult to completely avoid with automated algorithms, this invention innovatively integrates the IGV genome browser interface and establishes a dual verification system of initial screening and visual verification. The system directly projects the sequence alignment file and candidate mutation set to the verification interface, allowing for manual and intuitive inspection of the positive and negative strand distribution, end mismatches, and alignment quality values ​​of the reads. This effectively identifies and eliminates false positive sites caused by sequencing strand bias or read end artifacts. This human-computer interaction compensates for the blind spots of pure algorithm detection and provides strong quality assurance for generating the final high-confidence gene mutation detection report.

[0055] 4. This invention optimizes the analysis performance for tumor samples by calling the BWA alignment tool and the Vardict variant detection algorithm; in particular, it performs local re-alignment during the identification process, effectively correcting the arrangement errors of inserted and missing regions, thereby accurately calculating allele frequencies; combined with a dual-mode detection architecture that supports single and paired samples, this method can not only sensitively capture complex genomic variations, but also stably detect low-frequency mutations in highly heterogeneous tumor samples, providing a solid data foundation for accurate tumor subtyping and personalized medication guidance. Attached Figure Description

[0056] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0057] Figure 1 This is a flowchart of the method of the present invention.

[0058] Figure 2 This is a system structure diagram of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0060] Example 1:

[0061] Please see Figure 1 Web-based visualization and analysis methods for detecting tumor gene mutations using whole-exome sequencing include:

[0062] The system collects raw sequencing data uploaded by users and information on the reference genome version selected by the users through a web interface.

[0063] The system calls a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generates quality-controlled sequencing data and a visualized quality control report.

[0064] Based on the reference genome version information, the alignment algorithm is used to map the quality-controlled sequencing data to the corresponding reference genome to generate a sequence alignment file.

[0065] The mutation detection algorithm is invoked to scan the sequence alignment file, identify genomic mutation sites, and generate a raw mutation record file containing mutation information.

[0066] Biological information annotation is performed on the variant sites in the original variant record file using a preset variant annotation program and database, and the annotated variant sites are screened at multiple levels according to a preset filtering strategy to generate a candidate pathogenic mutation set.

[0067] The candidate pathogenic mutation set is projected onto the visual verification interface. In response to the received manual inspection and verification instructions, the site status in the candidate pathogenic mutation set is confirmed or removed, and the final gene mutation detection report is generated.

[0068] This embodiment provides a web-based method for visual analysis of tumor gene mutation detection using whole-exome sequencing. This method aims to address the problems of high threshold for bioinformatics analysis, low visualization level, and high false positive rate due to the lack of manual verification in existing processes.

[0069] The specific implementation steps are as follows:

[0070] Data acquisition steps: The system provides a file upload portal through a web interface; users upload raw sequencing data through a browser and select the reference genome version information from the drop-down menu; to ensure transmission stability, this embodiment uses fragmented upload and breakpoint resume technology;

[0071] Quality control steps: After receiving the data, the server calls the preset quality control program to perform quality assessment and cleaning on the raw sequencing data; this step generates clean post-quality control sequencing data by identifying low-quality bases and adapter sequences, and simultaneously generates a visualized quality control report containing multi-dimensional indicators for users to view instantly on the web page.

[0072] Sequence alignment steps: Based on the reference genome version information, the system automatically loads the corresponding index file, uses the alignment algorithm to map the quality-controlled sequencing data to the corresponding reference genome, and generates a sequence alignment file that records the position information of each read segment on the genome;

[0073] The mutation detection step involves calling a mutation detection algorithm to scan the sequence alignment file. This step aims to identify sites that are inconsistent with the reference genome from a large number of normal sequences, i.e., genomic mutation sites, and generate an original mutation record file containing information such as mutation coordinates, reference bases, and mutated bases.

[0074] Annotation and screening steps: Biological information annotation is performed on the variant sites in the original variant record file using a preset variant annotation program and database, assigning them attributes such as gene name and functional impact; according to the preset filtering strategy, the annotated variant sites are screened at multiple levels to remove background noise and non-pathogenic sites, generating a highly reliable candidate pathogenic mutation set.

[0075] Visual verification step: Project the candidate pathogenic mutation set onto the visual verification interface; this interface integrates the genome browser function, allowing users to intuitively view the sequencing details of the mutation sites; the system responds to the received manual check and verification instructions, confirms or removes the site status in the candidate pathogenic mutation set, and finally generates a gene mutation detection report containing confirmed mutation sites.

[0076] This embodiment connects the entire process of whole-exome sequencing analysis through a web-based visualization, greatly reducing the barrier to entry for clinicians or researchers. In particular, the introduction of a visual verification interface and a manual inspection and verification process effectively solves the industry pain point that pure algorithm analysis cannot completely eliminate sequencing artifacts, thereby improving the accuracy and clinical reference value of the final test report.

[0077] The steps involved in using a pre-defined quality control program to assess and clean the raw sequencing data, and to generate post-quality-controlled sequencing data and a visualized quality control report include:

[0078] The Fastp quality control algorithm was used to perform a single scan on the raw sequencing data;

[0079] Calculate the average read length, repetition rate, peak value of inserted fragments, total number of bases, Q20 base ratio, Q30 base ratio and GC content of the raw sequencing data to generate raw data quality control indicators;

[0080] Perform low-quality read filtering and adapter sequence removal operations to generate quality control sequencing data;

[0081] The quality control data indicators were calculated based on the post-quality control sequencing data, and the original data quality control indicators and the post-quality control data indicators were rendered into a visual quality control report in HTML and JavaScript object representation JSON format.

[0082] This embodiment is a further specification of the quality control steps;

[0083] The specific implementation process of calling the preset quality control program is as follows:

[0084] Algorithm selection and scanning: This embodiment uses the Fastp quality control algorithm to perform a single scan on the raw sequencing data; Fastp is an ultra-fast all-in-one preprocessing tool, whose advantage is that it does not require step-by-step processing, and data reading and index calculation can be completed in one I / O;

[0085] Indicator Calculation: During the scanning process, the following key indicators are calculated in parallel to generate raw data quality control indicators:

[0086] Average read length: reflects the uniformity of sequencing fragments;

[0087] Repeatability: Reflects the degree of redundancy introduced by PCR amplification;

[0088] Insertion fragment peaks: assessing library construction quality;

[0089] Total base count: measures whether the amount of sequencing data meets the depth requirements;

[0090] Q20 base ratio: The percentage of bases with a sequencing quality value greater than 20, i.e., an error rate of <1%;

[0091] Q30 base ratio: The percentage of bases with a sequencing quality value greater than 30, i.e., an error rate of <0.1%;

[0092] GC content: the ratio of guanine to cytosine, used to assess sequencing bias;

[0093] Cleaning operation: Based on preset thresholds, low-quality reads are filtered and adapter sequences are removed; for example, if the proportion of N bases in a read exceeds 10% or the average quality value is below 20, it is discarded, thereby generating clean post-quality control sequencing data.

[0094] Report rendering: The quality control data indicators are calculated based on the sequencing data after quality control. The backend encapsulates the original data quality control indicators and the data indicators after quality control into JSON format data. The frontend uses HTML and JavaScript object notation JSON format to parse the data and uses chart libraries such as ECharts to render it into a visual quality control report, which intuitively shows the quality comparison before and after filtering.

[0095] By employing the Fastp algorithm and combining it with JSON / HTML visualization rendering, this embodiment not only significantly improves the processing speed of massive sequencing data, but also enables users to quickly determine the usability of sequencing data through intuitive interactive reports, avoiding the waste of subsequent analysis resources due to poor raw data quality.

[0096] Based on reference genome version information, the steps for mapping post-quality control sequencing data to the corresponding reference genome and generating sequence alignment files using alignment algorithms include:

[0097] In response to the reference genome version information, load the corresponding reference genome index file;

[0098] The BWA alignment tool is used to align the quality-controlled sequencing data with the loaded reference genome to generate initial alignment results.

[0099] The initial alignment results are format-converted and sorted to generate a sequence alignment file in binary alignment format (BAM).

[0100] This embodiment is a further specification of the sequence alignment steps;

[0101] The specific process for generating sequence alignment files based on reference genome version information is as follows:

[0102] Index loading: In response to the reference genome version information, the system retrieves and loads the corresponding reference genome index file from the server file system based on the user's selection in the interface, such as GRCh37 or GRCh38.

[0103] Alignment execution: The BWA alignment tool is invoked to align the quality-controlled sequencing data with the loaded reference genome. The BWA-MEM algorithm utilizes the longest exact match strategy to efficiently handle long reads and gap alignments, generating initial alignment results that include alignment location, alignment quality, and CIGAR string.

[0104] Format processing: In order to optimize storage space and reading efficiency, the system calls the Samtools tool to perform format conversion and sorting of the initial alignment results; specifically, the text format SAM is converted into binary compressed format and sorted according to chromosome coordinate order to generate the final binary alignment format BAM sequence alignment file.

[0105] This embodiment achieves flexible support for multiple versions of the genome and efficient storage of large amounts of data by dynamically loading the index and generating binary BAM files. The sorted BAM files support random access, providing the necessary index foundation for subsequent variant detection and IGV visualization, and significantly improving the system's I / O performance.

[0106] The steps involved in using a variant detection algorithm to scan the sequence alignment file, identify genomic variant sites, and generate a raw variant record file containing variant information include:

[0107] Determine the variant detection mode, which may include single-sample mode or paired-sample mode;

[0108] Use the Vardict mutation detection algorithm to read the sequence alignment file;

[0109] The sequence alignment file identifies insertion / deletion variants, single nucleotide variants, and complex variants, and performs local realignment to correct allele frequencies.

[0110] The identified variant sites and their corresponding genotype information and sequencing depth information are written into the original variant record file of the VCF variant calling format;

[0111] This embodiment is a further specification of the mutation detection steps;

[0112] The specific implementation process of calling the variant detection algorithm to identify genomic variant sites is as follows:

[0113] Mode determination: The system first reads the user configuration to determine the mutation detection mode;

[0114] Single-sample mode: Only tumor tissue sequencing data is input, suitable for finding all non-reference sites;

[0115] Paired sample mode: Simultaneously input tumor tissue and normal control tissue to clearly mark the source attribute of the mutation in the original mutation record file through comparative analysis, thereby distinguishing germline variation from somatic mutation;

[0116] Algorithm call: Call the Vardict variant detection algorithm to read the sequence alignment file; Vardict is an ultrasensitive variant detector, especially suitable for detecting low-frequency mutations in whole-exome sequencing;

[0117] Variation identification and correction: The Vardict algorithm identifies insertion / deletion variants, single nucleotide variants, and complex variants in sequence alignment files; during this process, the system performs local re-alignment to correct allele frequencies;

[0118] Explanation of the necessity of local realignment: Because alignment algorithms are prone to mismatches when processing sequences near the Indel, resulting in false positive SNVs; local realignment corrects mismatches by reconstructing haplotypes in the Indel region, thereby obtaining more accurate mutation frequencies.

[0119] Output results: The system writes the identified variant sites and their corresponding genotype information, sequencing depth information, and variant allele frequencies into the original variant record file of the VCF variant calling format;

[0120] This embodiment employs the Vardict algorithm combined with a local re-alignment strategy, which significantly enhances the detection sensitivity for insertion / deletion variants and low-frequency somatic variants. This is particularly important for tumor samples, as tumor samples often exhibit heterogeneity and have a low mutation frequency, making them prone to missed by conventional algorithms. This solution effectively addresses this issue.

[0121] The steps for annotating biological information about variant sites in a raw variant record file using a pre-defined variant annotation program and database include:

[0122] Call the mutation annotation program to parse the original mutation record file;

[0123] The variant sites in the original variant record file are matched with preset frequency databases and functional databases; the frequency databases include the 1000 Genomes Database and the Exome Aggregation Consortium Database, and the functional databases include the COSMIC Tumor Mutation Database and the ClinVar Clinical Variance Database.

[0124] Add population frequency information, functional prediction information, and pathogenicity classification information to the successfully matched variant sites to generate an annotated variant list;

[0125] This embodiment is a further specification of the annotation step in the annotation screening process;

[0126] The specific implementation process of biological information annotation using a pre-defined variation annotation program and database is as follows:

[0127] Program call: Calls the mutation annotation program to parse the original mutation record file;

[0128] Database matching: The system matches the variant sites in the original variant record file with the preset frequency database and functional database;

[0129] Frequency databases include the 1000 Genomes Database and the Exome Aggregation Consortium Database; their function is to provide the frequency of occurrence of variant sites in healthy individuals, which is used for subsequent exclusion of polymorphic sites.

[0130] Functional databases include the COSMIC tumor mutation database, which provides information on known somatic mutations related to cancer, and the ClinVar clinical variant database, which provides the level of evidence for the relationship between variants and human diseases.

[0131] Information addition: Add population frequency information, functional prediction information, and pathogenicity classification information to the successfully matched variant sites, and finally generate an annotated variant list;

[0132] By integrating multi-dimensional biological databases, this embodiment transforms tedious variant sites containing only coordinate information into biologically meaningful and information-rich data. This enables subsequent screening strategies to be based not only on statistical indicators but also on biological functions and clinical evidence, significantly improving the clinical relevance of test results.

[0133] The steps for generating a candidate pathogenic mutation set by performing multi-level screening of annotated variant sites according to a preset filtering strategy include:

[0134] First-level filtering: Traverse the annotated mutation list, remove mutation sites whose frequency in the frequency database is higher than the preset background frequency threshold, and retain rare mutation sites;

[0135] Second-level screening: Examine the genomic regions and functional impacts of rare variant sites, retain sites located in exon regions or splice regions and whose mutation type is non-synonymous mutation, and generate a set of functionally relevant variants;

[0136] Level 3 screening: Obtain the sequencing support count and allele frequency of each site in the functionally relevant variant set, and perform the following classification and screening logic:

[0137] When the mutation type is determined to be germline mutation, the site with a sequencing support number greater than the preset first depth threshold and the mutation allele frequency greater than the preset first frequency threshold is retained, and the site is included in the candidate pathogenic mutation set.

[0138] When the mutation type is determined to be a somatic mutation, the site with a sequencing support number greater than the preset second depth threshold and the mutation allele frequency greater than the preset second frequency threshold is retained, and the site is included in the candidate pathogenic mutation set.

[0139] Sites that fail to meet the above conditions for sequencing support number or variant allele frequency are filtered out.

[0140] This embodiment is a further specification of the filtering step in the annotation filtering process;

[0141] The specific implementation process of multi-level screening of annotated variant sites based on a preset filtering strategy is as follows. This strategy is designed as a funnel-shaped three-level filtering model:

[0142] First-level screening:

[0143] Objective: To exclude common benign polymorphic sites in the population;

[0144] Operation: Traverse the annotated mutation list and remove those whose frequency in the frequency database exceeds a preset background frequency threshold. The mutation sites; in this embodiment, The value is set to 0.01. Any variant with a population frequency greater than 1% is considered a polymorphism and is removed, thus preserving rare variant sites.

[0145] Second-level screening:

[0146] Objective: To preserve variations that have a substantial impact on protein function;

[0147] Operation: Examine the genomic regions and functional impacts of rare variant sites; the system retains sites located in exon regions or splice regions, and whose mutation types are nonsynonymous mutations, frameshift mutations, or nonsense mutations that alter the protein sequence, while removing synonymous mutations and intron variants to generate a set of function-related variants;

[0148] Third-level screening:

[0149] Objective: To ensure the reliable sequencing quality of variants and to apply different confidence levels based on the source of the variant;

[0150] Obtain the number of sequencing supports for each site in the set of functionally relevant variants. and frequency of variant alleles The system reads the mutation detection mode parameters from the configuration file and executes the following classification and filtering logic:

[0151] Scenario A: If the current mode is paired samples, branch processing is performed based on the somatic cell state markers output by the Vardict algorithm:

[0152] When the variant type is determined to be germline variant, the number of sequencing supports is greater than the preset first depth threshold. And the frequency of the mutated allele is greater than the preset first frequency threshold. The site was identified and included in the candidate pathogenic mutation set;

[0153] When the mutation type is determined to be a somatic mutation, the number of sequencing supports must be greater than the preset second depth threshold. And the frequency of the variant allele is greater than the preset second frequency threshold. The site was identified and included in the candidate pathogenic mutation set;

[0154] Scenario B: If the current mode is single-sample, due to the lack of normal control samples, the system will by default mark all variant sites that have passed the first two levels of screening as suspected pathogenic variants and apply the preset universal screening threshold; it will retain those with a sequencing support number greater than the preset third depth threshold. Furthermore, the frequency of the mutated allele is greater than the preset third frequency threshold. The site was identified and included in the candidate pathogenic mutation set;

[0155] Parameter example: ;

[0156] Sites that fail to meet the above conditions for sequencing support number or variant allele frequency are filtered out.

[0157] The three-level cascaded filtering strategy designed in this embodiment progressively advances from population frequency and biological function to sequencing quality; in particular, it sets differentiated criteria for germline and somatic cell variations. and The threshold ensures the reliability of high-frequency germline mutations while avoiding the missed detection of low-frequency important driver gene mutations, achieving the best balance between sensitivity and specificity.

[0158] Projecting the candidate pathogenic mutation set onto the visual verification interface, and responding to received manual inspection and verification instructions, the steps for confirming or removing site status in the candidate pathogenic mutation set include:

[0159] Call the Genome Browser IGV interface to load sequence alignment files and candidate pathogenic mutation sets in the visualization verification interface;

[0160] The data shows the read alignment at the variant sites, including alignment quality values, positive and negative strand distribution, and read end position distribution.

[0161] In response to manual determination that the variant site has chain bias or is located in a high mismatch region at the end of a read, a removal instruction is received, the variant site is marked as a false positive and removed from the results;

[0162] In response to the manual determination that the alignment quality of the variant site is qualified and there are no obvious sequencing artifacts, a confirmation instruction is received, and the variant site is marked as the final pathogenic mutation.

[0163] This embodiment is a further specification of the visual verification steps;

[0164] The specific implementation process of projecting the candidate pathogenic mutation set onto the visual verification interface and confirming or eliminating it is as follows:

[0165] Interface call and loading: The front-end page initializes the igv.js component and obtains the access paths of the sequence alignment file and index file provided by the back-end through an HTTP request; the igv.js component reads the BAM file data through the HTTPRangeHeader segmented request mechanism according to the path, and renders the alignment view in real time on the visualization verification interface; at the same time, the front-end receives the coordinates of the candidate pathogenic mutation set transmitted by the back-end and controls the central window of the interface to automatically jump to the genomic coordinates of the first candidate mutation site;

[0166] Visualization: The system displays the read alignment results at the variant sites on the interface; to assist manual judgment, the following features are highlighted:

[0167] Compare quality values: MappingQuality is displayed through color depth or bar charts;

[0168] Distribution of positive and negative strands: This shows whether the mutation occurs simultaneously on both the positive and negative strands of DNA, or only on one strand.

[0169] Read end position distribution: The system calculates the relative position values ​​of variant bases on each supporting read. ,in This represents the relative offset of the variant base within the read segment. Given the total length of the read segment, count all segments that support reading. The values ​​are then used to generate a histogram of the distribution of read positions; if the histogram shows that the variant reads are concentrated in... or The range indicates the presence of a terminal effect; the above and The threshold setting is based on the principle of sequencing-by-synthesis in the second-generation sequencing platform, and is determined considering the statistical characteristic that the error rate increases exponentially when the enzyme activity decreases at both ends of the read during the sequencing reaction.

[0170] False positive rejection logic:

[0171] Strand bias: If a mutated read appears only on the positive strand or only on the negative strand, it usually indicates a sequencing error;

[0172] Terminal effect: If mutations are always located in high mismatch regions at the ends of reads, it is usually an alignment error;

[0173] Operation: In response to manual determination that the variant site has chain bias or is located in a high mismatch region at the end of a read, and upon receiving the removal instruction, the system backend immediately calls the database update interface to update the status field of the variant site record from candidate to false_positive, and records the operator ID and timestamp; at the same time, the front-end interface triggers a callback function to hide or mark the site as gray and unselectable from the current candidate pathogenic mutation set list, thereby marking the variant site as a false positive and removing it from the results;

[0174] True positive confirmation logic:

[0175] Operation: In response to manual determination that the alignment quality of the variant site is qualified and there are no obvious sequencing artifacts, a confirmation instruction is received, and the variant site is marked as the final pathogenic mutation;

[0176] The system records all manual operation logs and generates a final report based on them;

[0177] This embodiment introduces an IGV visualization interface to make the black-box algorithm results transparent; by manually reviewing features that are difficult to quantify perfectly by algorithms, such as chain bias and terminal distribution, it can effectively eliminate stubborn false positives in machine interpretation, so that the accuracy of the final report reaches the level of clinical application.

[0178] Example 2:

[0179] Please see Figure 2The data acquisition module is used to collect raw sequencing data uploaded by users and the reference genome version information selected by users through a web interactive interface.

[0180] The quality control module is configured to call a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generate quality-controlled sequencing data and a visual quality control report.

[0181] The sequence alignment module is configured to map the quality-controlled sequencing data to the corresponding reference genome based on the reference genome version information and use an alignment algorithm to generate a sequence alignment file.

[0182] The variant detection module is configured to call a variant detection algorithm to scan the sequence alignment file, identify genomic variant sites, and generate a raw variant record file containing variant information.

[0183] The annotation and filtering module is configured to use a preset variant annotation program and database to annotate the variant sites in the original variant record file with biological information, and to perform multi-level filtering on the annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set.

[0184] The visualization verification module is configured to project the candidate pathogenic mutation set onto the visualization verification interface, and in response to the received manual inspection and verification instructions, confirm or remove the site status in the candidate pathogenic mutation set, and generate the final gene mutation detection report.

[0185] This embodiment provides a web-based whole-exome sequencing tumor gene mutation detection visualization and analysis system. The system is based on a microservice architecture, with modules communicating via internal message queues. Specifically, it includes the following functional modules:

[0186] Data acquisition module: This module is responsible for file interaction between the front end and the user, and is used to collect raw sequencing data uploaded by the user and the reference genome version information selected by the user through the web interface;

[0187] Quality control module: This module integrates the Fastp engine and is configured to call preset quality control programs to perform quality assessment and cleaning of raw sequencing data, and generate quality-controlled sequencing data and visualized quality control reports.

[0188] Sequence alignment module: This module integrates BWA-MEM and Samtools toolchains, and is configured to map the quality-controlled sequencing data to the corresponding reference genome based on the reference genome version information using an alignment algorithm to generate sequence alignment files;

[0189] Variance detection module: This module serves as the core of the computation and is configured to call the variant detection algorithm to scan the sequence alignment file, identify genomic variant sites, and generate a raw variant record file containing variant information.

[0190] Annotation and filtering module: This module maintains a localized biological database and is configured to use a preset variant annotation program and database to annotate the variant sites in the original variant record file with biological information, and to perform multi-level filtering on the annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set.

[0191] Visual verification module: This module integrates the igv.js front-end component and is configured to project the candidate pathogenic mutation set onto the visual verification interface. In response to the received manual inspection and verification instructions, it confirms or removes the site status in the candidate pathogenic mutation set and generates the final gene mutation detection report.

[0192] The modular design of this system decouples the various functional units, making it easy to maintain and upgrade them independently. For example, when a new version of the reference genome or a new annotation database is released, only the corresponding alignment module or annotation module needs to be updated, without having to reconstruct the entire system, which has extremely high system scalability and maintenance efficiency.

[0193] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A Web-based method for visual analysis of tumor gene mutation detection using whole-exome sequencing, characterized in that, include: The system collects raw sequencing data uploaded by users and information on the reference genome version selected by the users through a web interface. The preset quality control program is invoked to perform quality assessment and cleaning on the raw sequencing data, generating quality-controlled sequencing data and a visual quality control report; Based on the reference genome version information, the quality control sequencing data is mapped to the corresponding reference genome using an alignment algorithm to generate a sequence alignment file. The mutation detection algorithm is invoked to scan the sequence alignment file, identify genomic mutation sites, and generate an original mutation record file containing mutation information. Biological information annotation is performed on the mutation sites in the original mutation record file using a preset mutation annotation program and database, and the annotated mutation sites are screened at multiple levels according to a preset filtering strategy to generate a candidate pathogenic mutation set. The candidate pathogenic mutation set is projected onto the visual verification interface. In response to the received manual inspection and verification command, the site status in the candidate pathogenic mutation set is confirmed or removed, and the final gene mutation detection report is generated.

2. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 1, characterized in that, The steps of calling a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generating quality-controlled sequencing data and a visualized quality control report include: The raw sequencing data were scanned in a single operation using the Fastp quality control algorithm. Calculate the average read length, repetition rate, peak value of inserted fragments, total number of bases, Q20 base ratio, Q30 base ratio and GC content of the raw sequencing data to generate raw data quality control indicators; Perform low-quality read filtering and adapter sequence removal operations to generate the quality control sequencing data; Based on the post-quality control sequencing data, the post-quality control data indicators are calculated, and the original data quality control indicators and the post-quality control data indicators are rendered into the visual quality control report in HTML and JavaScript object representation JSON format.

3. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 1, characterized in that, The step of mapping the quality-controlled sequencing data to the corresponding reference genome using an alignment algorithm based on the reference genome version information to generate a sequence alignment file includes: In response to the reference genome version information, the corresponding reference genome index file is loaded; The BWA alignment tool was invoked to align the quality-controlled sequencing data with the loaded reference genome, generating initial alignment results. The initial alignment results are format-converted and sorted to generate the sequence alignment file in binary alignment format (BAM).

4. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 1, characterized in that, The steps of calling the variant detection algorithm to perform a traversal scan of the sequence alignment file, identify genomic variant sites, and generate an original variant record file containing variant information include: Determine the variant detection mode, which includes a single-sample mode or a paired-sample mode; The Vardict mutation detection algorithm is invoked to read the sequence alignment file; Insertion / deletion variants, single nucleotide variants, and complex variants are identified in the sequence alignment file, and local re-alignment is performed to correct allele frequencies. The identified variant sites and their corresponding genotype information and sequencing depth information are written into the original variant record file of the VCF variant calling format.

5. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 1, characterized in that, The step of annotating the variant sites in the original variant record file with biological information using a preset variant annotation program and database includes: The mutation annotation program is invoked to parse the original mutation record file; The variant sites in the original variant record file are matched with preset frequency databases and functional databases; wherein, the frequency databases include the 1000 Genomes Database and the Exome Aggregation Consortium Database, and the functional databases include the COSMIC Tumor Mutation Database and the ClinVar Clinical Variance Database. Add population frequency information, functional prediction information, and pathogenicity classification information to the successfully matched variant sites to generate an annotated variant list.

6. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 5, characterized in that, The step of performing multi-level screening of annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set includes: First-level filtering: Traverse the variant list after the annotation, remove variant sites with frequencies higher than the preset background frequency threshold in the frequency database, and retain rare variant sites; Second-level screening: Examine the genomic regions and functional impacts of the rare variant sites, retain sites located in exon regions or splice regions and whose mutation type is non-synonymous mutation, and generate a set of functionally relevant variants; Third-level screening: Obtain the sequencing support number and variant allele frequency for each site in the functionally relevant variant set, and perform the following classification and screening logic: When the mutation type is determined to be germline mutation, the site with a sequencing support number greater than a preset first depth threshold and a mutation allele frequency greater than a preset first frequency threshold is retained, and the site is included in the candidate pathogenic mutation set. When the mutation type is determined to be a somatic mutation, the site with a sequencing support number greater than the preset second depth threshold and the mutation allele frequency greater than the preset second frequency threshold is retained, and the site is included in the candidate pathogenic mutation set. Sites that fail to meet the above sequencing support number or variant allele frequency conditions are filtered out.

7. The Web-based whole-exome sequencing tumor gene mutation detection visualization analysis method as described in claim 6, characterized in that, The step of projecting the candidate pathogenic mutation set onto the visual verification interface and, in response to a received manual inspection and verification command, confirming or removing the site status in the candidate pathogenic mutation set includes: Call the Genome Browser IGV interface to load the sequence alignment file and the candidate pathogenic mutation set in the visualization verification interface; The alignment of reads at the mutation sites is displayed, including alignment quality values, positive and negative strand distribution, and read end position distribution. In response to a manual determination that the variant site has chain bias or is located in a high mismatch region at the end of a read, a removal instruction is received, the variant site is marked as a false positive and removed from the results; In response to manual determination that the alignment quality of the variant site is qualified and there are no obvious sequencing artifacts, a confirmation instruction is received, and the variant site is marked as the final pathogenic mutation.

8. A web-based whole-exome sequencing tumor gene mutation detection visualization analysis system, based on the web-based whole-exome sequencing tumor gene mutation detection visualization analysis method according to any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect raw sequencing data uploaded by users and the reference genome version information selected by users through a web interactive interface; The quality control module is configured to call a preset quality control program to perform quality assessment and cleaning on the raw sequencing data, and generate quality-controlled sequencing data and a visual quality control report. The sequence alignment module is configured to map the quality control sequencing data to the corresponding reference genome based on the reference genome version information using an alignment algorithm, and generate a sequence alignment file. The variant detection module is configured to call a variant detection algorithm to perform a traversal scan of the sequence alignment file, identify genomic variant sites, and generate an original variant record file containing variant information. The annotation and filtering module is configured to use a preset variant annotation program and database to annotate the variant sites in the original variant record file with biological information, and to perform multi-level filtering on the annotated variant sites according to a preset filtering strategy to generate a candidate pathogenic mutation set. The visualization verification module is configured to project the candidate pathogenic mutation set onto the visualization verification interface, and in response to the received manual inspection and verification command, to confirm or remove the site status in the candidate pathogenic mutation set and generate the final gene mutation detection report.

Images