A method for determining and verifying the source of amino acid substitutions based on pan-modified protein variants, genomic and transcriptomic variant evidence, secondary spectra

Abstract

The application discloses a method for determining and verifying the source of amino acid substitution based on pan-modified protein variants, genomic and transcriptomic variation evidence and secondary spectra, belonging to the fields of pan-modified proteomics, genomics, transcriptomics, mass spectrometry data processing and computer-aided quality control. The method obtains pan-modified protein variants or substitution PSM data, a reference protein database, transcript annotation, genomic VCF, transcriptomic variation or expression evidence, a public variation database and MS / MS secondary spectrum information; standardizes amino acid substitution events and maps them from peptide coordinates to protein coordinates, transcript CDS coordinates and genomic coordinates; performs quality gating in combination with genomic and transcriptomic variation evidence; constructs a competitive set of substitution peptides, reference peptides, homologous peptides and modified peptides / quality offset explanations, and performs competitive verification based on secondary spectra; and finally fuses variation support, secondary spectrum support, modification conflicts, database uniqueness and boundary condition evidence to output P_variant, P_tx_tl, P_artifact, P_uncertain and other source support values, source classification, evidence abstracts and recommended review methods.

Description

Technical Field

[0001] This invention relates to the fields of panmodification proteomics, bioinformatics, genomics, transcriptomics, secondary mass spectrometry data processing, and computer-aided quality control. Specifically, it relates to a method for determining and verifying the source and confidence level of amino acid substitutions by integrating protein variants or amino acid substitution events identified in panmodification proteomics with evidence such as genomic variation in samples, transcriptomic variation or expression evidence, public variant databases, competitive matching of substitution peptides / reference peptides, and fragment ion consistency in MS / MS secondary spectra. This method can be implemented by corresponding systems, electronic devices, and computer-readable storage media.

[0002] This invention can be applied to panmodified proteomics, open-search proteomics, proteome-genome-transcriptome joint analysis, tumor mutant peptide validation, personalized protein database construction, amino acid substitution false positive filtering, and automated interpretation of customer reports. Background Technology

[0003] (a) Current state of technology In panmodification proteomics or open-search proteomics analyses, search engines may report amino acid substitution peptides or quality shift events. These events may originate from genuine DNA variations, RNA editing or transcriptional errors, translational misincorporation, differences in reference databases or species annotations, homologous protein differences, post-modification mismatches, indistinguishable isomers / isomeric residues, and low-quality spectral matching errors.

[0004] Genomics and transcriptomics analyses can generate sample-specific VCF files using WES, WGS, or RNA-seq, recording genomic coordinates, reference bases, substitution bases, sequencing depth, genotype quality, variant allele frequencies, and functional annotations. Public databases such as dbSNP, gnomAD, and COSMIC can provide evidence of population- or tumor-related variants.

[0005] Existing proteogenomics workflows typically identify variant peptides using personalized protein databases, two-pass search, open search, sample-specific variant peptide databases, and FDR controls. However, identifying a substitute peptide does not equate to confirming the source of the substitution; without consideration of the substitute peptide competing with the reference peptide for matching, localizing ions, retention time, fragment intensity, consistency, and external validation, a high false positive rate may still occur.

[0006] (ii) Problems with existing technology 1. The source classification is too broad. Current procedures often interpret replacements as "mutant peptides" or "quality shifts," but fail to adequately differentiate between DNA variant support, RNA / translation errors, PTM or database mismatches, random spectral errors, and uncertain sources.

[0007] 2. Relying solely on PSM score or FDR is insufficient to confirm the source of a substitution. Substitute peptides may have scores close to those of reference peptides, homologous peptides, or modified peptides, which can lead to poor interpretation conflicts, especially in general modification or open searches.

[0008] 3. Version and transcriptomic issues exist in the mapping from protein coordinates to genome coordinates. Different FASTA, GTF / GFF, reference genome, and transcriptomic versions may lead to inconsistencies in coordinates. Multiple transcriptomic and isoform versions can introduce multiple candidate mapping paths.

[0009] 4. Conventional MS / MS have natural boundaries for some cases of equal or near-equal mass. For example, Leu / Ile is usually indistinguishable under normal collision-induced fragmentation conditions. Without setting exclusion or uncertainty markers, common-sense errors are easily generated.

[0010] 5. Existing results often lack VCF quality thresholds, class-specific FDR, RT / fragmentation prediction consistency, multiple read coverage, and synthetic peptide / relabeled peptide validation suggestions, making it difficult to support high-confidence alternative sources conclusions. Summary of the Invention

[0011] (a) Purpose of the invention The purpose of this invention is to provide a method for determining and verifying the source of amino acid substitutions based on panmodified protein variants, genomic and transcriptomic variation evidence, and secondary spectra. This method addresses the problems in existing proteomics analyses, such as unclear sources of protein variants or substitution peptides, complex sources of false positives, unstable coordinate mapping, insufficient evaluation of secondary spectra evidence, and lack of boundary conditions and interpretable evidence.

[0012] This invention outputs four or more sources of evidence, evidence levels, and suggested review methods through substitution event standardization, protein-transcript-genome multipath mapping, sample genome VCF and transcriptome variation or expression evidence matching, public database matching, secondary spectrum competition verification of substitution peptide / reference peptide / homologous peptide / modified peptide, retention time and fragment ion consistency assessment, equal quality case labeling, two-stage or class-specific FDR control, and multi-evidence source probability fusion.

[0013] (II) Technical Solution 1. Overall Structure Overview. The method can be executed by a system, which includes a data input and version checking module, a panmodified protein variant extraction and normalization module, a protein-transcript-genome coordinate mapping module, a genome and transcriptome variant matching and quality gating module, a secondary spectrogram competition verification module, an equal quality and modification conflict determination module, a source probability fusion module, and a result output module.

[0014] 2. Data Input and Version Check Module. This module receives panmodified protein variants or substitution PSM results, protein FASTA, reference genome version, GTF / GFF or CDS annotation files, sample genome VCF, optional RNA-seq variants or expression files, public variant databases, databases of common poor-quality PTMs, and raw or parsed MS / MS secondary spectrogram information. The system checks the version consistency between FASTA, GTF / GFF, reference genome, VCF, and transcriptome annotations; if inconsistencies are found, coordinate transformation is performed or a version inconsistency risk marker is output.

[0015] 3. Substitution Site Extraction and Standardization Module. This module standardizes the substitution representations output by different search software into the format of "reference amino acid-protein position-substitute amino acid", such as E123K, and retains the peptide's local location, protein ID, isoform, secondary spectrum ID, search task ID, PSM score, q-value, PEP, FDR, and original mass offset. For peptides with multiple substitutions, the system establishes event records for each and retains the correlations between events.

[0016] 4. Protein-Transcript-Genome Coordinate Mapping Module. This module maps substitution events from local peptide locations to protein coordinates, transcript CDS coordinates, and genome coordinates based on protein sequences, CDS annotations, and exon coordinates. If multiple transcripts or proteins match, the system retains candidate paths and generates path confidence based on the main transcript, expression level, protein evidence, annotation level, peptide uniqueness, and coordinate consistency.

[0017] 5. Variance Matching and Quality Gating Module. This module matches candidate genome coordinates and inferred codon changes with sample VCFs and public databases. Sample VCFs are preferred to meet preset quality thresholds, such as sequencing depth (DP), genotype quality (GQ), variant allele frequency (VAF), and PASS markers; specific thresholds can be configured according to WES, WGS, or RNA-seq data types. Public databases can be used as supplementary evidence but must not replace sample-specific variant evidence.

[0018] 6. Secondary Spectral Competition Validation Module. This module constructs a competitive matching set for substitution peptides, reference peptides, homologous candidate peptides, and potentially modified peptides. Δscore is defined as the difference between the best matching score of the substitution peptide and the best non-substitution interpretation score in the competitive set, rather than comparing it only with a single reference peptide. The system simultaneously calculates the number of localizing ions, b / y ion coverage on both sides of the substitution site, precursor ion mass error, fragment ion intensity consistency, isotope rationality, retention time prediction bias, and duplicate detection.

[0019] 7. Equivalent Quality and Modification Conflict Determination Module. For equivalent quality residues such as Leu / Ile that are difficult to distinguish using conventional MS / MS, the system defaults to outputting a "equivalent quality indistinguishable" or "origin uncertain" label, unless supported by a special fragmentation pattern, chemical derivatization, or orthogonal experiments. For substitutions with quality differences approaching common PTMs, the system searches the PTM quality difference database and evaluates the competition scores between modification and substitution explanations.

[0020] 8. Source Probability Fusion Module. This module integrates features such as variant support, VCF quality, public database support, transcript expression, Δscore, localizing ions, RT / fragment consistency, duplicate detection, multiple read coverage, modification conflict, and database uniqueness, outputting at least four types of source support values: P_variant indicates DNA or sample variant-supported substitution; P_tx_tl indicates RNA / transcriptional error or translational misincorporation-supported substitution; P_artifact indicates modification, database, or search interpretation conflict-supported substitution; P_uncertain indicates random error, insufficient evidence, or substitution that cannot be distinguished by current data.

[0021] 9. Results Output Module. Output fields include protein ID, gene name, substitution type, protein location, transcript ID, genomic coordinates, candidate codon changes, VCF support information, DP / GQ / VAF, public database ID, Δscore, number of localizing ions, RT bias, fragment consistency, multiple read coverage, P_variant, P_tx_tl, P_artifact, P_uncertain, final classification, level of evidence, and suggested review method.

[0022] (III) Beneficial Effects 1. This invention maps substitution events from proteomics results to transcript and genomic coordinates, and combines VCF quality indicators to enhance the feasibility and verifiability of substitution source interpretation.

[0023] 2. This invention uses competitive matching of alternative peptides, reference peptides, homologous peptides, and modified peptides to define Δscore, avoiding logical loopholes caused by simply comparing alternative peptides with a single reference peptide.

[0024] 3. This invention introduces boundary condition markers for equal-mass residues, near-equal-mass modifications, and low-coverage sites to avoid drawing overly generalized conclusions about situations that are difficult to distinguish using existing mass spectrometry techniques.

[0025] 4. This invention reduces false positives for substitution peptides and improves reporting reliability through two-stage or class-specific FDR, RT / fragment consistency, multiple reads coverage, and orthogonal validation recommendations.

[0026] 5. This invention outputs support values ​​and evidence summaries from multiple sources, which can reduce the workload of manual verification and is suitable for high-throughput generalized modified proteomics or proteogenomics technical service processes. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall process of a method for determining and verifying the source of amino acid substitutions based on panmodified protein variants, genomic and transcriptomic variation evidence, and secondary spectra. It shows the process of inputting substitution PSM results, reference protein database, transcript annotation, sample VCF, common variant database, and original secondary spectra information into the system, followed by substitution site extraction and standardization, protein-transcript-genome coordinate mapping, variant matching and quality gating, secondary spectra competitive verification, equal quality / modification conflict determination, and source probability fusion, and finally outputting source classification, evidence summary, and suggested review methods.

[0028] Figure 2 This is a schematic diagram of substitution site extraction and protein-transcript-genome coordinate mapping in this invention. It shows the process of identifying standardized events (such as E123K) by identifying amino acid substitutions in peptide segments, and mapping the local position of the peptide segment to protein coordinates, transcript CDS coordinates and genome coordinates step by step. When there are multiple transcripts or multiple mapping paths, the system retains the path confidence and outputs the standardized multipath mapping results.

[0029] Figure 3 This is a schematic diagram of variant matching and secondary spectrum competition verification in this invention. It shows the process of matching candidate genome coordinates with sample VCF, transcriptome variant or expression evidence, and public variant database to generate variant supporting evidence. It also shows the process of MS / MS secondary spectrum competition matching of the competitive interpretation set consisting of substitution peptide, reference peptide, homologous peptide, and modified peptide / mass shift interpretation. The credibility of the substitution source is evaluated based on indicators such as Δscore, localization ion, ion coverage, mass error, RT bias, and duplicate detection consistency.

[0030] Figure 4 This is a schematic diagram of the multi-evidence source probability fusion and classification output in this invention. It shows that after the evidence input source probability fusion model is used to evaluate VCF quality, public database support, Δscore, localized ions, RT / fragment consistency, modification conflict, database uniqueness, and multiple reads coverage, it outputs source support values ​​such as P_variant, P_tx_tl, P_artifact, and P_uncertain. It further forms classification results such as variant support type, transcription / translation related candidate type, modification / database conflict type, and insufficient evidence or source uncertainty type. At the same time, it outputs evidence summary and suggested review methods. Detailed Implementation

[0031] (I) Specific structural implementation In one specific embodiment, the present invention is implemented by a server-side bioinformatics analysis pipeline. The server configuration includes a processor, memory, storage, and optional parallel computing resources. The storage holds a reference protein database, a reference genome, transcript CDS annotations, a common variant database, a PTM poor quality database, retention time / fragmentation prediction models, project configuration files, and output report templates.

[0032] The system input files include: a substitution PSM result table, a protein FASTA file, a GTF / GFF or CDS annotation file, a sample VCF file, a public variant library index file, raw spectra or analyzed secondary spectra information, and optional RNA-seq expression matrix or mutation table. The system output includes a substitution source determination table, a detailed evidence table, quality control statistics, and a report text.

[0033] Preferably, the system employs a two-stage search or class-specific FDR control: the first stage uses a conventional reference library to identify high-confidence conventional peptides and removes interpreted spectra; the second stage searches the remaining spectra or candidate replacement libraries and performs individual FDR or posterior error rate control on the replacement peptide category.

[0034] (II) Implementation of the work process Step 1: Import search results from Open Search, Generalized Modification Search, or Individualized Database, extract Replacement PSMs, and standardize the event representation. The system records the original quality offset and its possible PTM interpretations.

[0035] Step two involves mapping the replacement site from its local peptide location to full-length protein coordinates, and then further mapping it to transcript and genome coordinates based on CDS annotation. If multiple candidate transcripts exist, the system retains multiple paths and calculates path confidence.

[0036] Step 3: Match candidate coordinates with the sample VCF and check for reference bases, substituted bases, codon changes, amino acid alterations, DP, GQ, VAF, and PASS markers. If the sample VCF is missing, the system only outputs auxiliary evidence from the public database and does not consider it as support for high-confidence sample variation.

[0037] Step four involves constructing a competitive interpretation set of substitution peptides, reference peptides, homologous peptides, and potentially modified peptides. The system calculates Δscore, localizing ion, ion coverage, RT bias, fragment intensity consistency, and repeat detection; when substitutions fall under indistinguishable quality cases such as Leu / Ile, the system outputs boundary condition markers.

[0038] Step 5: Input all evidence into the fusion model and output P_variant, P_tx_tl, P_artifact, and P_uncertain. For events with acceptable VCF quality, high Δscore, sufficient localization ions, consistent RT / fragmentation, and multiple read coverage, the system classifies them as "variant-supported replacement"; for events without sample VCF support, with higher competition scores for modification interpretations, or insufficient localization evidence, they are classified as "modification / database / search interpretation conflict"; for events of equal quality and indistinguishable or with insufficient evidence, they are classified as "uncertain source".

[0039] (III) Boundary Conditions and Alternative Implementation Methods When no sample VCF or RNA-seq data is available, this invention can still output a preliminary classification based on public databases and secondary spectroscopic evidence, but it should not be listed as high-confidence sample variant support.

[0040] When there are conflicts in Leu / Ile, Gln / Lys approximate quality, poor quality of modification and substitution, or low fragmentation coverage, the system output may be indistinguishable or have an uncertain source. Orthogonal validation using special fragmentation, synthetic peptide, relabeled peptide, or targeted mass spectrometry is recommended.

[0041] When the reference database and genome annotation versions are inconsistent, the system can complete the analysis by using version conversion, rebuilding the individualized database, or retaining multiple mapping paths, and mark the version risk in the results.

[0042] Source probability fusion models can employ logistic regression, Bayesian networks, random forests, XGBoost, support vector machines, neural networks, or rule-based models, but all should retain variation quality gating, competitive spectrogram verification, boundary condition labeling, and evidence summary output.

[0043] (iv) Data Fields, Quality Gating and Module Implementation Details 1. Data Input Fields. Substitution-type PSM input fields may include sample number, batch number, secondary spectrum number, peptide sequence, protein number, gene name, amino acid before substitution, amino acid after substitution, peptide local location, protein location, search engine score, q-value, PEP, FDR, precursor ion quality error, fragment ion matching number, peptide intensity, retention time, and original quality offset. Reference data input fields may include reference protein FASTA version, reference genome version, GTF / GFF annotation version, transcript ID, CDS start and end positions, exon coordinates, protein-transcript correspondence, common variant database version, and common PTM quality difference database version. Sample VCF input fields may include chromosome, genome coordinates, reference bases, substitution bases, genotype, sequencing depth (DP), genotype quality (GQ), variant allele frequency (VAF), filter status (PASS), functional annotation, transcript annotation, and amino acid change annotation.

[0044] 2. Data Preprocessing Steps. The system standardizes the substitution representations output by different search software, normalizing substitution events to the form of "reference amino acid-protein position-substitute amino acid," while retaining the original software output, secondary spectrum numbers, search task numbers, and quality offset values. The system checks version consistency among FASTA, GTF / GFF, reference genome, and VCF, removes low-quality PSMs, and establishes candidate path records for multi-protein matching, multi-transcript mapping, and homologous protein matching. For cases lacking sample VCF or RNA-seq, the system can still perform preliminary classification, but will not output high-confidence sample variant support conclusions.

[0045] 3. Quality Gating Rules. Mass spectrometry-level gating includes FDR threshold, q-value threshold, minimum spectral matching score, precursor ion mass error range, fragment ion coverage requirements, replacement site localization ion requirements, retention time deviation range, and repeat detection requirements at the PSM or peptide level. Variation-level gating includes DP, GQ, VAF, PASS status, reference base consistency, codon change consistency, and functional annotation consistency of the sample VCF. Annotation-level gating includes transcript annotation completeness, protein-transcriptograph mapping uniqueness, public database version reliability, and database uniqueness. Events that fail to meet quality gating are marked as low confidence or of uncertain origin.

[0046] 4. Coordinate Mapping and Path Confidence. The protein-transcript-genome coordinate mapping module maps the replacement site from local peptide coordinates to full-length protein coordinates, and then maps it to transcript and genome coordinates based on CDS annotation. If multiple transcripts, isoforms, or proteins match, the system retains multiple candidate mapping paths and generates path confidence based on the main transcript, expression level, protein evidence, annotation level, peptide uniqueness, splicing consistency, and coordinate consistency.

[0047] 5. Secondary Spectral Competition Validation and Feature Construction. The secondary spectral competition validation module constructs a competitive interpretation set of substitution peptides, reference peptides, homologous peptides, and potentially modified peptides. Δscore is defined as the difference between the best matching score of the substitution peptide and the best non-substitution interpretation score in the competitive set. The system simultaneously evaluates the number of ions locating substitution sites, b / y ion coverage, parent ion mass error, fragment ion intensity consistency, retention time deviation, isotope rationality, and repeat detection consistency. Feature construction includes variant-supporting features, spectral-supporting features, modification conflict features, database uniqueness features, repeat detection features, and boundary condition features.

[0048] 6. Fusion Logic. The source probability fusion module inputs features such as VCF quality, public database support, Δscore, localized ions, ion coverage, RT / fragment consistency, modification conflict, database uniqueness, multiple read coverage, and path confidence into the fusion model, and outputs P_variant, P_tx_tl, P_artifact, and P_uncertain. Here, P_variant represents a DNA or sample variant supporting substitution, P_tx_tl represents candidates related to RNA editing, transcriptional errors, or translational misincorporation, P_artifact represents conflicting explanations related to modifications, database annotations, or quality shifts, and P_uncertain represents insufficient evidence or uncertain source.

[0049] 7. Output Results. Output fields include protein ID, gene name, substitution event, protein location, transcript ID, genomic coordinates, candidate codon changes, VCF support information, DP / GQ / VAF, public database ID, Δscore, number of localizing ions, retention time bias, fragment consistency, multiple read coverage, P_variant, P_tx_tl, P_artifact, P_uncertain, final source classification, boundary condition markers, evidence summary, and suggested review method. The system can generate structured tables for reporting and detailed evidence for manual review.

[0050] 8. Boundary Conditions. The system outputs uncertain sources or suggests manual verification and labeling for cases such as missing sample VCF, missing RNA-seq data, inconsistent reference database versions, multiple transcripts not uniquely mapped, missing localizing ions at substitution sites, indistinguishable Leu / Ile residues, poor substitution quality closely resembling common PTMs, low fragmentation coverage, low repeat detection, and low VCF quality. For events requiring high-confidence conclusions, the system may suggest orthogonal validation using synthetic peptides, relabeled peptides, targeted mass spectrometry, special fragmentation methods, or database updates.

[0051] 9. Alternative Implementation Methods. Coordinate mapping can employ strategies prioritizing primary transcripts, expression levels, annotation confidence, or multi-path retention; spectral validation can utilize probability scoring, ion matching scores, machine learning spectral quality scoring, or manual spectral review instead of a single Δscore; source probability fusion can employ logistic regression, Bayesian networks, random forests, XGBoost, support vector machines, neural networks, or rule-based models; in the absence of sample VCF, a public variant database and secondary spectral evidence can be used for low-confidence initial screening.

Claims

1. A computer-implemented method for determining and verifying the origin of amino acid substitutions based on panmodified protein variants, genomic and transcriptomic variation evidence, and secondary spectra, characterized in that, The steps include: obtaining PSM data of panmodified protein variants or those containing amino acid substitutions, reference protein databases, transcript annotation data, sample genomic variation data, transcriptome variation or expression evidence, and MS / MS secondary spectrogram information; Amino acid substitution events are extracted and normalized from the PSM data; these events are mapped from peptide coordinates to protein coordinates, transcript coordinates, and genomic coordinates, and mapping path confidence is generated; the genomic coordinates are matched with sample VCF, transcriptome variation or expression evidence, and public variation databases, and quality gating is performed to generate variation supporting evidence; a competitive matching set of substitution peptides, reference peptides, homologous peptides, and potentially modified peptides is constructed, and secondary spectroscopic supporting evidence is generated based on the MS / MS secondary spectroscopic data; boundary condition markers are generated based on equal-quality residues, modification quality differences, and database versions. The mutation support evidence, secondary spectrum support evidence, boundary condition labeling, and modification conflict evidence are input into the source probability fusion model, which outputs the amino acid substitution source classification, source support value, evidence summary, and suggested review method.

2. The method according to claim 1, characterized in that, The generalized modified protein variants or PSM data containing amino acid substitutions include one or more of the following: sample number, batch number, secondary spectrum number, peptide sequence, protein number, gene name, amino acid before substitution, amino acid after substitution, peptide local position, protein position, search engine score, q-value, posterior error probability, FDR, precursor ion mass error, fragment ion matching number, peptide intensity, retention time, and original mass shift. The amino acid substitution event is normalized to a combination of reference amino acid, protein position, and substituted amino acid, while retaining the original mass shift, secondary spectrum number, and search task number.

3. The method according to claim 1, characterized in that, The mapping between protein coordinates, transcript coordinates, and genome coordinates includes converting protein coordinates to transcript coordinates based on CDS annotation, and then converting them to genome coordinates. When there are multiple transcripts, multiple isoforms, or multiple protein matches, multiple candidate mapping paths are retained, and path confidence is generated based on one or more of the following: main transcript, expression level, protein evidence, annotation level, peptide uniqueness, and coordinate consistency.

4. The method according to claim 1, characterized in that, The variant support evidence includes one or more of the following: sample VCF support, public database support, reference base consistency, codon change consistency, variant allele frequency (VAF), sequencing depth (DP), genotype quality (GQ), and PASS markers; the variant quality gating is performed based on one or more of the following: DP, GQ, VAF, PASS status, public database match confidence, and transcript annotation integrity.

5. The method according to claim 1, characterized in that, In the secondary spectrum competitive verification, Δscore is the difference between the best matching score of the substitution peptide and the best non-substitution interpretation score in the competitive matching set; The supporting evidence from the secondary spectra includes one or more of the following: number of ions at substitution sites, b / y ion coverage, parent ion mass error, consistency of fragment ion intensity, retention time prediction bias, reasonableness of isotopes, and consistency of detection in repeated experiments.

6. The method according to claim 1, characterized in that, The boundary condition markers include indistinguishable markers of equal quality, modification conflict markers, version risk markers, or markers of uncertain origin; when there are equal quality residue substitutions such as Leu / Ile that are difficult to distinguish by conventional MS / MS, substitution quality that is very close to that of common PTM, low fragment coverage, low duplicate detection, low VCF quality, or inconsistency with the reference database version, the corresponding boundary condition markers are generated.

7. The method according to claim 1, characterized in that, The source support values ​​include DNA or sample variation support values ​​P_variant, RNA editing, transcriptional errors or translational misincorporation related candidate support values ​​P_tx_tl, modification or database interpretation conflict support values ​​P_artifact, and insufficient evidence or source uncertainty support values ​​P_uncertain; the source probability fusion model is any one of logistic regression, Bayesian network, random forest, XGBoost, support vector machine, neural network or rule model.

8. The method according to claim 1, characterized in that, The output also includes one or more of the following: protein ID, gene name, substitution event, protein location, transcript ID, genomic coordinates, candidate codon changes, VCF support information, DP / GQ / VAF, public database ID, Δscore, number of localizing ions, RT bias, fragment consistency, multiple read coverage, boundary condition markers, and final source classification; the suggested verification methods include one or more of the following: synthetic peptide verification, targeted mass spectrometry verification, manual spectral verification, or database update.

9. A system for implementing the method for determining and verifying the source of amino acid substitutions according to any one of claims 1 to 8, characterized in that, The method includes a data input and version checking module, a panmodified protein variant extraction and standardization module, a protein-transcript-genome coordinate mapping module, a genome and transcriptome variant matching and quality gating module, a secondary spectrum competition verification module, an equal quality and modification conflict determination module, a source probability fusion module, and a result output module. Each module is configured to perform the method described in any one of claims 1 to 8.

10. An electronic device or computer-readable storage medium, characterized in that, The electronic device includes a processor and a memory, or the computer-readable storage medium stores a computer program; when the computer program is executed by the processor, it implements the method according to any one of claims 1 to 8.

Images