Genetic variation interpretation method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a genetic variation interpretation method and device, computer equipment, a readable storage medium and a program product, and the method comprises the steps: obtaining to-be-interpreted genetic data, determining the multi-modal data of a reference corresponding to at least one variation site in the to-be-interpreted genetic data based on a pre-constructed search database, and obtaining the multi-modal data of the reference corresponding to at least one variation site in the to-be-interpreted genetic data; and according to the at least one variation site and the multi-modal data of the corresponding reference, determining a corresponding variation rating evidence and a support basis of the variation rating evidence, and based on an upgrading and degrading strategy corresponding to the variation rating evidence, carrying out upgrading and degrading processing on the variation rating evidence according to the support basis of the variation rating evidence. And obtaining a variation rating result. According to the method, full automation of ACMG literature evidence rating is achieved by automatically reading literature content, and the working efficiency of genetic variation interpretation can be remarkably improved; and the accuracy and the reliability of the final variation rating result can be remarkably improved by performing upgrading and downgrading processing on the variation rating evidence.

Description

Technical Field

[0001] The present application relates to the field of bioinformatics, and in particular to a method, apparatus, computer device, readable storage medium and program product for interpreting genetic variation. Background Art

[0002] In genetic diagnosis and genomic research, the pathogenicity assessment of variants is a crucial step. According to the guidelines of the American College of Medical Genetics and Genomics (ACMG), the pathogenicity assessment of genetic variants requires extensive literature support, especially evidence involving novel variants (PS2, PM6), functional experiments (PS3, BS3), high-incidence variants in patient populations (PS4), and co-segregation in families (PP1, BS4).

[0003] In traditional techniques, genetic counselors must screen for variant evidence by carefully reading and analyzing text, images, tables, and other information in a large amount of medical literature. The traditional ACMG literature evidence method relies on manual screening of literature and analyzing the text descriptions, tables, and images related to variants in the literature one by one. Although this method helped genetic counselors in the early stages, with the surge in the number of genetic report interpretations, the method of obtaining evidence through manual reading of literature can no longer meet the growing diagnostic needs. It is not only time-consuming and labor-intensive, but also susceptible to human errors, thus affecting accuracy. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product that can efficiently and accurately interpret genetic variations in response to the above-mentioned technical problems.

[0005] In a first aspect, the present application provides a method for interpreting genetic variation, comprising:

[0006] Acquiring genetic data to be interpreted, wherein the genetic data to be interpreted includes at least one variant site;

[0007] Determining, based on a pre-constructed search database, multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted;

[0008] Determining corresponding variation rating evidence and supporting basis for the variation rating evidence according to the at least one variation site and the multimodal data of the corresponding references;

[0009] Based on the upgrading and downgrading strategy corresponding to the variation rating evidence, the variation rating evidence is upgraded and downgraded according to the supporting basis of the variation rating evidence to obtain a variation rating result.

[0010] In one embodiment, the method further includes: extracting multimodal data related to the variation from medical literature, the multimodal data including at least one of text, images and tables; constructing a search database based on the extracted multimodal data related to the variation, the search database including an association between variation information and the multimodal data.

[0011] In one of the embodiments, the corresponding variant rating evidence and the supporting basis of the variant rating evidence are determined by using a trained multimodal document evidence recognition model; the multimodal document evidence recognition model is trained by the following method: multimodal data related to the variant is extracted from medical literature, and the multimodal data includes at least one of text, image and table; based on the document evidence evaluation standard defined by ACMG, the multimodal data is evidence labeled to obtain a multimodal evidence dataset; based on the multimodal evidence dataset, the LoRA fine-tuning method is used to train the selected target multimodal large language model to obtain the multimodal document evidence recognition model.

[0012] In one of the embodiments, the method of determining the corresponding variant rating evidence and the supporting basis of the variant rating evidence by using the trained multimodal document evidence recognition model includes: generating evidence recognition prompt information based on the multimodal data of the at least one variant site and the corresponding reference document; inputting the evidence recognition prompt information into the multimodal document evidence recognition model to obtain the evidence recognition result output by the multimodal document evidence recognition model, and the evidence recognition result includes the variant rating evidence and the supporting basis of the variant rating evidence.

[0013] In one of the embodiments, the upgrading and downgrading strategy corresponding to the variant rating evidence is used to upgrade and downgrade the variant rating evidence according to the supporting basis of the variant rating evidence to obtain a variant rating result, including: generating upgrading and downgrading prompt information based on the variant rating evidence, the supporting basis of the variant rating evidence and the upgrading and downgrading strategy corresponding to the variant rating evidence; inputting the upgrading and downgrading prompt information into the multimodal document evidence recognition model to obtain the variant rating result output by the multimodal document evidence recognition model.

[0014] In one of the embodiments, extracting multimodal data related to the variation from the medical literature includes: parsing the medical literature to convert the medical literature into a structured multimodal data format; and extracting multimodal data related to the variation from the structured multimodal data.

[0015] In a second aspect, the present application also provides a genetic variation interpretation device, the device comprising:

[0016] An acquisition module, used for acquiring genetic data to be interpreted, wherein the genetic data to be interpreted includes at least one variation site;

[0017] A first determination module is used to determine multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted based on a pre-built search database;

[0018] A second determination module is used to determine corresponding variation rating evidence and supporting basis of the variation rating evidence according to the at least one variation site and the multimodal data of the corresponding reference;

[0019] The result output module is used to upgrade or downgrade the variation rating evidence according to the supporting basis of the variation rating evidence based on the upgrade or downgrade strategy corresponding to the variation rating evidence to obtain the variation rating result.

[0020] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described in the first aspect when executing the computer program.

[0021] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0022] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0023] The above-mentioned genetic variation interpretation method, device, computer equipment, computer-readable storage medium and computer program product obtain the genetic data to be interpreted, determine the multimodal data of the reference corresponding to at least one variation site in the genetic data to be interpreted based on the pre-constructed search database, determine the corresponding variation rating evidence and the supporting basis of the variation rating evidence based on at least one variation site and the corresponding multimodal data of the reference, and upgrade and downgrade the variation rating evidence based on the upgrading and downgrading strategy corresponding to the variation rating evidence and the supporting basis of the variation rating evidence to obtain the variation rating result. It can realize the full automation of ACMG literature evidence rating by automatically reading the content of the literature, which can significantly improve the work efficiency of genetic variation interpretation; and by upgrading and downgrading the variation rating evidence, it can significantly improve the accuracy and reliability of the final variation rating result. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments of the present application or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 is a schematic diagram of a process for interpreting genetic variation in one embodiment;

[0026] Figure 2 A schematic diagram of a flow chart of a model training step in an embodiment;

[0027] Figure 3 A schematic diagram of a flow chart of the upgrade and downgrade processing steps in one embodiment;

[0028] Figure 4 is a structural block diagram of a genetic variation interpretation device in one embodiment;

[0029] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0031] The traditional method of manually reading literature to obtain evidence for interpreting genetic reports can no longer meet the growing diagnostic needs, is time-consuming and labor-intensive, and is easily affected by human errors and accuracy.

[0032] Although there are tools such as InterVar (a bioinformatics software tool) and VarSome (a powerful genetic variation annotation and analysis tool) that can realize the automated analysis of some non-documentary evidence, manual judgment is still required for documentary evidence. Most of the tools that can realize the automatic rating of ACMG evidence items focus on automating the prediction of some non-documentary evidence from the perspective of bioinformatics analysis, while the evaluation of documentary evidence still mainly relies on manual reading and interpretation. Although some studies have proposed the idea of ​​automatic extraction of documentary evidence, these methods are usually limited to building a local documentary evidence library to extract documentary evidence of some interpreted variants, and have not yet fully realized the automated rating of documentary evidence. At the same time, traditional methods lack in-depth understanding and analysis of multimodal data such as text and charts. Therefore, traditional tools still find it difficult to accurately realize the fully automatic identification of documentary evidence.

[0033] Based on this, the embodiment of the present application provides a method for interpreting genetic variation, which fully utilizes the advantages of multimodal large model technology and combines the ACMG literature evidence rules to automatically identify multimodal evidence information such as text, pictures and tables from medical literature. It can also identify and analyze pictures (such as family charts, functional experimental results charts) and tables (such as case data, statistical results tables) in the literature, thereby comprehensively integrating various forms of evidence information in the literature and improving the accuracy and efficiency of ACMG literature evidence interpretation in genetic interpretation.

[0034] In one embodiment, Figure 1 As shown, a method for interpreting genetic variation is provided, the method comprising the following steps:

[0035] Step 102, obtaining genetic data to be interpreted.

[0036] Among them, genetic data is an important basis for studying biological genetic variation, gene function and evolution, and covers biomedical data with multiple levels of information such as genome, transcriptome, and epigenome. In this embodiment, the genetic data to be interpreted includes at least one mutation site. Specifically, the mutation site refers to the specific location where the mutation occurs in the genome, usually one or more bases in the genome sequence have changed. These mutations may occur at any position in the DNA (DeoxyriboNucleicAcid, deoxyribonucleic acid) sequence, including coding regions (such as inside genes) and non-coding regions (such as intergenic regions, regulatory elements, etc.). Mutation sites are important concepts in genetics and genomics research because they may affect the function, expression and regulation of genes, thereby affecting the characteristics of organisms or causing diseases.

[0037] Specifically, when interpreting genetic variation, it is first necessary to obtain the genetic data to be interpreted, and the genetic data to be interpreted includes at least one variant site, and then through subsequent steps to analyze and interpret it, it can reveal the dynamic changes of the genome and its impact on biological traits. For example, some variant sites may be related to genetic diseases or complex diseases. By analyzing the variant sites, personalized treatment plans can be developed.

[0038] Step 104 , based on a pre-constructed search database, determine multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted.

[0039] The search database is pre-built based on medical references and includes associations between variant information and multimodal data in the literature. Variant information refers to entities such as genes, variants, and diseases that are valuable for judging ACMG literature evidence. Multimodal data includes at least one of text, images, and tables related to variant information in the literature.

[0040] In this embodiment, by pre-building a search database, after obtaining the genetic data to be interpreted based on the above steps, the multimodal data of the reference corresponding to at least one variant site in the genetic data to be interpreted can be determined based on the pre-built search database. Specifically, the search database can be searched based on at least one variant site to query the multimodal data of the reference that has an associated relationship with the at least one variant site.

[0041] Step 106, determining corresponding variation rating evidence and supporting basis of the variation rating evidence based on at least one variation site and the multimodal data of the corresponding reference.

[0042] Among them, variant rating evidence refers to literature evidence that complies with ACMG guidelines, including but not limited to benign evidence items (such as BS3, BS4, BP2, BP5, etc.) and pathogenic evidence items (such as PS2, PS3, PS4, PM3, PM6, PP1, PP4, etc.). Supporting evidence can be key information such as the number of probands, functional experimental results, and new variants related to variant rating evidence extracted and summarized from the literature.

[0043] Specifically, based on at least one variant site and the multimodal data of the corresponding references, the corresponding variant rating evidence and the supporting basis of the variant rating evidence can be determined through analysis and evaluation.

[0044] Step 108, based on the upgrade and downgrade strategy corresponding to the variation rating evidence, the variation rating evidence is upgraded and downgraded according to the supporting basis of the variation rating evidence to obtain the variation rating result.

[0045] The upgrade and downgrade strategy can be a standard for upgrading and downgrading literature evidence items formulated in accordance with the ACMG guidelines based on the characteristics of different variant rating evidence. The variant rating result refers to the final evaluation result of the genetic data to be interpreted after upgrading and downgrading the variant rating evidence.

[0046] In this embodiment, based on the upgrading and downgrading strategies corresponding to the variation rating evidence, the variation rating evidence determined in the above steps is upgraded and downgraded according to the supporting basis of the variation rating evidence, so as to obtain a more accurate variation rating result.

[0047] In the above-mentioned method for interpreting genetic variation, by obtaining the genetic data to be interpreted, based on the pre-constructed search database, the multimodal data of the reference corresponding to at least one variant site in the genetic data to be interpreted is determined, and the corresponding variant rating evidence and the supporting basis of the variant rating evidence are determined according to the at least one variant site and the corresponding multimodal data of the reference, and based on the upgrading and downgrading strategy corresponding to the variant rating evidence, the variant rating evidence is upgraded and downgraded according to the supporting basis of the variant rating evidence to obtain the variant rating result. It can significantly improve the work efficiency of genetic variation interpretation by automatically reading the content of the literature, thereby realizing the full automation of ACMG literature evidence rating; and by upgrading and downgrading the variant rating evidence, the accuracy and reliability of the final variant rating result can be significantly improved.

[0048] In one embodiment, in step 106, determining the corresponding variant rating evidence and the supporting basis of the variant rating evidence may specifically include: determining the corresponding variant rating evidence and the supporting basis of the variant rating evidence by using a trained multimodal literature evidence recognition model. The multimodal literature evidence recognition model may be an artificial intelligence model that is trained to accurately identify ACMG literature evidence items related to variants (including variant rating evidence and corresponding supporting basis).

[0049] In this embodiment, evidence identification prompt information can be generated based on at least one variation site in the genetic data to be interpreted and the corresponding multimodal data of the reference literature; and the evidence identification prompt information is input into the multimodal literature evidence identification model to obtain the evidence identification result output by the multimodal literature evidence identification model, wherein the evidence identification result includes variation rating evidence and supporting basis for the variation rating evidence.

[0050] Specifically, the evidence identification prompt information may be a relevant prompt for instructing the multimodal literature evidence identification model to perform literature evidence identification, and may specifically include a task goal prompt. For example, the task goal prompt may be in a similar form to "according to the input variant information and related literature information, quickly and accurately identify the type of literature evidence supported by the variant in accordance with the ACMG guidelines." Among them, the specific content of "variation information" may be at least one variant site in the input genetic data to be interpreted, and at least one of the HGVS (Human Genome Variation Society) expression, gene alias, and gene-related disease information of the variant site. The specific content of "related literature information" may be multimodal data of references corresponding to at least one variant site, such as related text fragments and charts.

[0051] In one scenario, the evidence identification prompt information may also include a defined output format prompt. For example, the output format prompt may be in a similar form as "If the document supports multiple evidence types, return a list of evidence rating types and their reasons that meet the ACMG standard in json format; if there is no supporting evidence, return an empty json{}". This enables the model to output evidence identification results that meet the requirements.

[0052] In addition, the evidence identification prompt information can also include defined task prompts. For example, the task prompt can be a similar form of "Since PS2 has a higher priority, if it is rated as PS2, PM6 does not need to be rated. If there is not enough evidence in the literature to support any evidence type of the ACMG standard, an empty json{} is returned. For parts involving family information, mutation frequency, etc., it is necessary to count the number of probands, the number of patients in the family, and other data." This can enhance the model's reasoning and understanding capabilities, thereby improving the model's processing efficiency and accuracy.

[0053] In an exemplary embodiment, Figure 2 As shown in the figure, the above multimodal document evidence recognition model can be trained by the following method:

[0054] Step 202, extract multimodal data related to the variation from medical literature.

[0055] Among them, medical literature is information related to medicine and has reference value. Multimodal data includes at least one of text, image and table in the literature.

[0056] In this embodiment, the medical literature may be parsed and converted into a structured multimodal data format; and then multimodal data related to the variation may be extracted from the structured multimodal data.

[0057] Specifically, medical literature (such as literature in PDF format) can be deeply analyzed, and text content such as titles, abstracts, keywords, text, references, supplementary materials, and non-text elements such as tables and pictures in the literature can be converted into data structures that are convenient for subsequent analysis. In this process, the original structure of the literature is retained, and the tables are also saved as pictures, and an accurate relationship index between the text and charts (pictures and tables) is established. It is also possible to accurately identify genes, variants, and disease entities in the literature based on natural language processing entity extraction technology, and standardize these entities through entity alignment technology to obtain extracted multimodal data related to variants.

[0058] In one scenario, in order to address the problem of information overload that may arise during the storage and use of huge amounts of information in medical literature, a binary classification model can be used to screen text fragments that are valuable for evidence judgment, thereby eliminating article fragments that cannot be used as ACMG literature evidence items and retaining only text fragments that are valuable for evidence judgment.

[0059] In another scenario, the variation information such as genes, variants, disease entities contained in each text fragment, as well as the associations between related multimodal data such as text fragments, images and table index information, can be stored in the local knowledge graph or Elasticsearch (search and analysis engine) database to obtain a constructed search database, laying a solid foundation for subsequent evidence rating work.

[0060] Step 204 , based on the literature evidence evaluation standard defined by ACMG, evidence annotation is performed on the multimodal data to obtain a multimodal evidence dataset.

[0061] Since finding documentary evidence items that meet the ACMG guidelines is a complex and meticulous task. Evidence is not only implicit in text fragments, but sometimes it is necessary to combine tables, pictures (such as experimental results, pedigree charts) and other multimodal information for comprehensive analysis to accurately extract specific documentary evidence items. If you only rely on the text content itself, it may not be enough to fully confirm specific evidence items in the document. In addition, other relevant evidence items may be omitted. Therefore, in this embodiment, based on the documentary evidence evaluation criteria defined by ACMG, multimodal data can be annotated for evidence, thereby obtaining a multimodal evidence dataset, and then the multimodal evidence dataset is used for model training, so as to enhance the comprehensiveness and accuracy of the model in the process of documentary evidence rating.

[0062] Among them, the ACMG-defined literature evidence evaluation criteria can be in the form shown in Table 1 below, including pathogenic evidence items such as PS2, PS3, PS4, PM3, PM6, PP1, and PP4, as well as benign evidence items such as BS3, BS4, BP2, and BP5. By annotating multimodal evidence data, more accurate support can be provided for the evaluation of literature evidence. Table 1 is:

[0063]

[0064] In this embodiment, for each variation that meets the ACMG literature evidence rating, the corresponding literature fragments and charts can be annotated with evidence items (such as PS2, PM3, PP1, etc. and their supporting evidence). Specifically, the following two methods can be used to collect and annotate evidence data:

[0065] (1) Annotation based on the accumulation of the internal report interpretation system. Since the internal report interpretation system records the literature fragments, tables, and pictures that genetic counselors have interpreted in the past and used as references for their variant evidence ratings, the corresponding literature fragments, tables, and pictures can be annotated as evidence items based on this record.

[0066] (2) Annotation based on the assistance of the big model. For example, by designing prompts that meet the ACMG evidence rating standards, the language comprehension ability of the big model can be used to analyze the text and chart content of the variant and related literature, obtain possible evidence fragments and charts, and annotate them.

[0067] In one scenario, the above-mentioned annotation content may also be reviewed and corrected to further improve the accuracy of the annotation content.

[0068] In one scenario, in order to better manage and annotate multimodal data, the annotated multimodal evidence data can also be centrally managed and stored. The storage content can include mapping relationships between variant information, evidence items (i.e., variant rating evidence and supporting evidence), reference IDs, document fragments (i.e., text fragments in the corresponding documents), reference table links, and reference image links. The table content can be stored in image format and saved to the local database together with the relevant images.

[0069] In one scenario, if the same document content or the same figure may involve multiple evidence items, multiple labels can be used for annotation. For example, for a document fragment with a document ID of "24664454", since it involves multiple evidence items (i.e., variant rating evidence and supporting evidence) such as PS2 and PS4, its evidence items can be annotated based on multiple labels such as PS2 and PS4.

[0070] In an exemplary embodiment, in addition to marking the evidence items in the literature that meet the ACMG guidelines, in order to improve the model's understanding of the chart information in the evidence rating data, some corresponding descriptions of the charts can also be marked. In particular, the relevant information of the pedigree chart, functional experiment chart, and variation table can enhance the model's in-depth understanding of the charts that meet the evidence rating standards. For example, the understanding of the offspring relationship, inheritance pattern, and disease status of the variant carrier in the pedigree chart can be marked; the understanding of the changes in protein expression levels and enzyme activity changes in the functional experiment chart can be marked, etc. It can be understood that this part marks the descriptive sentences corresponding to the charts in the literature that are meaningful for evidence rating, rather than evidence items (such as PS2, PP1, etc.).

[0071] Step 206: Based on the multimodal evidence dataset, the LoRA fine-tuning method is used to train the selected target multimodal large language model to obtain a multimodal document evidence recognition model.

[0072] In this embodiment, the model training process may include multimodal instruction fine-tuning data design, target multimodal large language model screening during fine-tuning training, fine-tuning strategy selection, etc.

[0073] Among them, the multimodal instruction fine-tuning data design can be combined with the rating standards of PS2, PS3, PS4, PM3, PM6, PP1, PP4, BS3, BS4, BP2, and BP5 evidence items in the ACMG guidelines, and the multimodal evidence instructions (that is, evidence identification prompt information) can be written in a chain of thought. Since gene-related diseases are of great reference significance in the rating process, for example, for PP4 evidence items, it is necessary to judge the high consistency between the patient's phenotype and the disease phenotype. Therefore, the prompt information should include disease information related to gene mutations, so as to judge whether the phenotype or family history of the case in the literature is highly consistent with the phenotype of the disease.

[0074] In the process of model training, two types of data, multimodal evidence data set and plain text evidence data set, can be used as training data and test data, that is, both training data and test data include these two types of data. Among them, multimodal evidence data includes text evidence data and related chart information, while plain text evidence data does not include chart information. In the above data collection and annotation process, in addition to collecting multimodal document evidence data, chart description data is also specially annotated. And the chart description data can be organized in a format similar to the multimodal evidence instruction data, thereby ensuring the consistency of the data structure.

[0075] During the model training process, first, you can select multiple pre-trained multimodal large language models as base models based on the local hardware environment and computing resources. These models can analyze natural language text information and images at the same time, understand multimodal data integrated with images and text, and have leading pure text understanding capabilities. For example, you can use the Qwen-VL series model (a large-scale visual language model) or Mistral AI (a company focusing on the field of artificial intelligence)'s multimodal Pixtral series model (a multimodal language model released by Mistral AI).

[0076] Secondly, the basic models are screened. Specifically, the test data of the above-organized instruction data can be input into multiple basic models for prediction, and the prediction results can be statistically analyzed to select the basic model that performs well in the ACMG literature evidence rating task as the target multimodal large language model to ensure its better understanding of the task. This not only helps to avoid poor fine-tuning effects due to large differences in pre-training and fine-tuning data, but also accelerates the convergence speed of model fine-tuning.

[0077] Then, an efficient fine-tuning strategy is selected. Since full parameter fine-tuning requires high computing resources and is time-consuming to train, this embodiment adopts the Lora (Low-Rank Adaptation, an efficient fine-tuning technology for pre-trained models) fine-tuning method. The Lora fine-tuning method only needs to update a small number of parameters, thereby reducing computing resource consumption while maintaining most of the performance of the pre-trained model.

[0078] Finally, based on the test data of multimodal instructions containing multimodal evidence instruction data and chart description instruction data, the above-selected target multimodal large language model was trained by multi-task fine-tuning to obtain a multimodal literature evidence recognition model. This multimodal multi-task fine-tuning method not only significantly improves the model's ability to understand various charts in the literature (especially family charts, variant functional experimental charts, and variant-related tables), but also ensures that the model can accurately identify ACMG literature evidence items related to variants. Therefore, the multimodal literature evidence recognition model obtained after training can be used to interpret actual genetic data, providing an efficient and accurate solution for the interpretation of genetic variants.

[0079] In an exemplary embodiment, Figure 3 As shown, in step 108, based on the upgrading and downgrading strategy corresponding to the variation rating evidence, the variation rating evidence is upgraded and downgraded according to the supporting basis of the variation rating evidence to obtain the variation rating result, which may specifically include:

[0080] Step 302: Generate upgrade and downgrade processing prompt information based on the variation rating evidence, the supporting basis of the variation rating evidence, and the upgrade and downgrade strategy corresponding to the variation rating evidence.

[0081] Among them, the upgrade and downgrade processing prompt information is similar to the above-mentioned evidence identification prompt information, and is a relevant prompt used to instruct the multimodal document evidence identification model to perform upgrade and downgrade processing. It can specifically include upgrade and downgrade instructions. For example, the upgrade and downgrade instructions can include a similar form of "You are an ACMG variant interpretation expert, who can upgrade and downgrade the variant evidence type according to the ACMG evidence upgrade and downgrade strategy and evidence basis, and output the upgrade and downgrade results." Among them, "evidence basis" includes variant rating evidence and corresponding supporting basis, which can be obtained by the fine-tuned multimodal document evidence identification model. For example, the evidence basis can be specifically in the form of "Evidence basis: PS4 document PMID 22464770 reported a proband with cardiomyopathy. PS4 document PMID 24503780 reported a proband with cardiomyopathy. PS4 document PMID27532257 reported a case of cardiomyopathy. PS4 document PMID 30996762 reported 3 siblings with isolated biventricular non-pathogenic cardiomyopathy. PS4 document PMID 28855170 reported a case of cardiomyopathy. PS4 document PMID 24664454 reported 2 probands with cardiomyopathy."

[0082] "Upgrading and downgrading strategy" refers to the description of the upgrade and downgrading of the corresponding evidence items specified by ACMG. For example, the upgrade and downgrading strategy for the evidence item PS4 (the frequency of this variant in the diseased population is significantly higher than that in the control population. Extremely rare variants may not be statistically significant in case-control studies. When the variant is observed in multiple probands with the same phenotype, different levels of PS4 evidence can be given according to the number of probands). The specific form can be "If there are 15 or more probands carrying the variant to be evaluated, PS4 will be assigned; if there are 6 or more probands carrying the variant to be evaluated but less than 15, PS4_Moderate will be assigned; if there are 2 or more probands carrying the variant to be evaluated but less than 6, PS4_Supporting will be assigned. Note: only the number of probands is counted, and the number of probands in the literature is not counted repeatedly."

[0083] In this embodiment, corresponding promotion and demotion processing prompt information can be generated based on the variant rating evidence, the supporting basis of the variant rating evidence, and the promotion and demotion strategy corresponding to the variant rating evidence. Usually, based on internal business needs, the rating results for benign evidence (including BS3, BS4, BP2, BP5) and PP4 can maintain the original ratings unchanged, while the rating results for PS2, PS3, PS4, PM3, PM6, PP1 can be promoted and demoted based on the corresponding promotion and demotion strategy, thereby achieving a comprehensive and accurate assessment of the pathogenicity of the variant.

[0084] Step 304: input the upgrade and downgrade processing prompt information into the multimodal document evidence recognition model to obtain the variation rating result output by the multimodal document evidence recognition model.

[0085] Among them, the multimodal literature evidence recognition model is obtained after the above-mentioned fine-tuning training. The model can not only accurately extract evidence items related to the variation, but also intelligently analyze and summarize the supporting basis of the evidence items. For example, for the PS4 evidence item, the model can automatically extract and summarize key information such as the number of probands. Based on the input variation information and the variation rating evidence and its supporting basis output by the model, combined with the ACMG guidelines, the corresponding literature evidence item upgrade and downgrade strategies formulated for the characteristics of each evidence item of PS2, PS3, PS4, PM3, PM6, and PP1 can generate corresponding upgrade and downgrade processing prompt information. By inputting the upgrade and downgrade processing prompt information into the multimodal literature evidence recognition model, the variation rating result output by the model can be obtained. It can be understood that the upgrade and downgrade processing prompt information can drive the multimodal literature evidence recognition model to realize the upgrade and downgrade processing of literature evidence, thereby significantly improving the accuracy and reliability of the pathogenicity assessment of the variation.

[0086] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0087] Based on the same inventive concept, the embodiment of the present application also provides a genetic variation interpretation device for implementing the genetic variation interpretation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more genetic variation interpretation device embodiments provided below can refer to the limitations of the genetic variation interpretation method above, and will not be repeated here.

[0088] In an exemplary embodiment, Figure 4 As shown, a genetic variation interpretation device is provided, comprising: an acquisition module 402, a first determination module 404, a second determination module 406 and a result output module 408, wherein:

[0089] An acquisition module 402 is used to acquire genetic data to be interpreted, wherein the genetic data to be interpreted includes at least one variation site;

[0090] A first determination module 404 is used to determine multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted based on a pre-built search database;

[0091] A second determination module 406 is used to determine corresponding variation rating evidence and supporting basis of the variation rating evidence according to the at least one variation site and the corresponding multimodal data of the reference document;

[0092] The result output module 408 is used to upgrade or downgrade the variation rating evidence according to the supporting basis of the variation rating evidence based on the upgrade or downgrade strategy corresponding to the variation rating evidence to obtain a variation rating result.

[0093] In an exemplary embodiment, the device also includes a search database construction module for extracting multimodal data related to the variation from medical literature, wherein the multimodal data includes at least one of text, images, and tables; based on the extracted multimodal data related to the variation, a search database is constructed, wherein the search database includes an association relationship between the variation information and the multimodal data.

[0094] In an exemplary embodiment, the second determination module may be specifically configured to determine corresponding variation rating evidence and supporting basis of the variation rating evidence by using a trained multimodal document evidence recognition model.

[0095] In an exemplary embodiment, the multimodal document evidence recognition model is trained by the following method: extracting multimodal data related to variations from medical literature, the multimodal data including at least one of text, images and tables; based on the literature evidence evaluation criteria defined by ACMG, annotating the multimodal data to obtain a multimodal evidence dataset; based on the multimodal evidence dataset, using the LoRA fine-tuning method to train the selected target multimodal large language model to obtain the multimodal document evidence recognition model.

[0096] In an exemplary embodiment, the second determination module can also be used to: generate evidence identification prompt information based on the multimodal data of the at least one variation site and the corresponding reference document; input the evidence identification prompt information into the multimodal document evidence identification model to obtain the evidence identification result output by the multimodal document evidence identification model, and the evidence identification result includes the variation rating evidence and the supporting basis for the variation rating evidence.

[0097] In an exemplary embodiment, the result output module is also used to: generate upgrade and downgrade processing prompt information based on the variation rating evidence, the supporting basis of the variation rating evidence, and the upgrade and downgrade strategy corresponding to the variation rating evidence; input the upgrade and downgrade processing prompt information into the multimodal document evidence recognition model to obtain the variation rating result output by the multimodal document evidence recognition model.

[0098] In an exemplary embodiment, the device also includes a data extraction module, which is used to: parse the medical literature and convert the medical literature into a structured multimodal data format; and extract multimodal data related to the variation from the structured multimodal data.

[0099] Each module in the above-mentioned genetic variation interpretation device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0100] In an exemplary embodiment, a computer device is provided, the internal structure of which can be shown as follows: Figure 5 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (Near Field Communication, NFC) or other technologies. When the computer program is executed by the processor, a method for interpreting genetic variation is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0101] Those skilled in the art will understand that Figure 5The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0102] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0103] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0104] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0105] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0106] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0107] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0108] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for interpreting genetic variation, characterized in that: The method comprises: Acquiring genetic data to be interpreted, wherein the genetic data to be interpreted includes at least one variant site; Determining, based on a pre-constructed search database, multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted; Determining corresponding variation rating evidence and supporting basis for the variation rating evidence according to the at least one variation site and the multimodal data of the corresponding references; Based on the upgrading and downgrading strategy corresponding to the variation rating evidence, the variation rating evidence is upgraded and downgraded according to the supporting basis of the variation rating evidence to obtain a variation rating result.

2. The method according to claim 1, characterized in that The method further comprises: Extracting multimodal data related to the variation from medical literature, wherein the multimodal data includes at least one of text, image, and table; Based on the extracted multimodal data related to the variation, a search database is constructed, wherein the search database includes an association relationship between variation information and the multimodal data.

3. The method according to claim 1, characterized in that Determine corresponding variant rating evidence and supporting basis of the variant rating evidence by using the trained multimodal literature evidence recognition model; The multimodal document evidence recognition model is trained by the following method: Extracting multimodal data related to the variation from medical literature, wherein the multimodal data includes at least one of text, image, and table; Based on the literature evidence evaluation standard defined by ACMG, the multimodal data is annotated to obtain a multimodal evidence dataset; Based on the multimodal evidence dataset, the LoRA fine-tuning method is used to train the selected target multimodal large language model to obtain the multimodal document evidence recognition model.

4. The method according to claim 3, characterized in that The method of determining the corresponding variant rating evidence and the supporting basis of the variant rating evidence by using the trained multimodal document evidence recognition model includes: Generate evidence identification prompt information according to the multimodal data of the at least one variant site and the corresponding reference; The evidence identification prompt information is input into the multimodal document evidence identification model to obtain an evidence identification result output by the multimodal document evidence identification model, wherein the evidence identification result includes the variation rating evidence and supporting basis for the variation rating evidence.

5. The method according to claim 3, characterized in that: The upgrading and downgrading strategy corresponding to the variation rating evidence is used to upgrade and downgrade the variation rating evidence according to the supporting basis of the variation rating evidence to obtain the variation rating result, including: Generate upgrade and downgrade processing prompt information based on the variation rating evidence, the supporting basis of the variation rating evidence, and the upgrade and downgrade strategy corresponding to the variation rating evidence; The upgrade and downgrade processing prompt information is input into the multimodal document evidence recognition model to obtain the variation rating result output by the multimodal document evidence recognition model.

6. The method according to any one of claims 2 to 5, characterized in that: The method of extracting multimodal data related to the variation from medical literature includes: Parsing the medical literature to convert the medical literature into a structured multimodal data format; Extracting variation-related multimodal data from structured multimodal data.

7. A genetic variation interpretation device, characterized in that: The device comprises: An acquisition module, used for acquiring genetic data to be interpreted, wherein the genetic data to be interpreted includes at least one variation site; A first determination module is used to determine multimodal data of references corresponding to at least one variant site in the genetic data to be interpreted based on a pre-built search database; A second determination module is used to determine corresponding variation rating evidence and supporting basis of the variation rating evidence according to the at least one variation site and the multimodal data of the corresponding reference; The result output module is used to upgrade or downgrade the variation rating evidence according to the supporting basis of the variation rating evidence based on the upgrade or downgrade strategy corresponding to the variation rating evidence to obtain the variation rating result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images