Gene prediction system based on phenotype-gene relevance and storage medium

Abstract

The invention discloses a gene prediction system based on phenotype-gene relevance and a storage medium, and the system comprises a data integration module, a knowledge graph construction module, a neural network module and a rapid updating module. The data integration module is used for integrating a plurality of medical databases to obtain a multi-modal database; the data of the multi-modal database is in a gene-disease-phenotype association form; the knowledge graph construction module is used for constructing a knowledge graph according to the data of the multi-modal database; the neural network module is used for processing the preprocessed data according to a knowledge graph and a multi-modal database to obtain a pathogenic gene prediction result; and the rapid updating module is used for updating the multi-modal database, the knowledge graph and the neural network module. According to the application, a plurality of medical databases are integrated through the data integration module to obtain the multi-modal database, and the preprocessed data are processed through the neural network module to obtain the pathogenic gene prediction result. Through correlation of phenotypes and genes, the accuracy of gene prediction is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically to a gene prediction system and storage medium based on phenotype-gene association. Background Art

[0002] Rare diseases usually have the characteristics of low incidence, few cases, complex and diverse clinical manifestations, and often involving multiple system involvements. It is estimated that there are 7,000 - 8,000 known rare diseases globally, accounting for about 10% of human diseases. However, due to their rarity, the patient population for each disease is relatively small, which makes traditional disease diagnosis and gene mapping methods based on large-scale population studies face great challenges in rare disease research. Many rare disease patients are difficult to be accurately diagnosed in the early stage of the disease and often experience multiple misdiagnoses and long medical treatment processes, which not only bring great physical and mental pain and economic burden to patients and their families, but also are not conducive to the effective treatment and management of the disease.

[0003] Currently, AI technology has been deeply integrated into the field of rare disease diagnosis, significantly improving the diagnosis efficiency through phenotype-gene association analysis. With the continuous deepening of clinical research and the development of medical informatization, more and more phenotype data of rare disease patients have been recorded and accumulated. These phenotype data include multiple aspects such as patients' symptoms, signs, laboratory test results, and imaging manifestations.

[0004] In order to better utilize these phenotype data, it is necessary to standardize and normalize them so that computers can identify and analyze them. Currently, some international organizations and research institutions are committed to establishing standard terms and databases for rare disease phenotypes, such as the Human Phenotype Ontology (HPO), etc., providing a standardized data basis for phenotype-gene association modeling.

[0005] However, currently the usage rate of HPO terms in clinical practice is less than 40%, and the conversion error rate between free text and structured terms can reach 25%, seriously affecting the quality of model input. Existing training sets overly rely on European population data (such as non-European samples in gnomAD only account for 17%), resulting in a 30 - 50% decrease in diagnostic sensitivity for the East Asian, especially the Han ethnic group. Summary of the Invention

[0006] This application is proposed to solve the above problems. According to one aspect of this application, a gene prediction system based on phenotype-gene association is provided, including a data integration module, a knowledge graph construction module, a neural network module, and a rapid update module:

[0007] The data integration module is used to integrate multiple medical databases to obtain a multi-modal database, and the data of the multi-modal database is in the form of gene-disease-phenotype association;

[0008] The knowledge graph construction module is used to construct a knowledge graph according to the data in the multimodal database;

[0009] The neural network module is used to process the preprocessed data based on the knowledge graph and the multimodal database to obtain a pathogenic gene prediction result;

[0010] The fast update module is used to update the multimodal database, the knowledge graph and the neural network module.

[0011] In an embodiment of the present application, the data integration module is used to integrate multiple medical databases to obtain a multimodal database, including:

[0012] The multiple medical databases include a first database, a second database, a third database, a fourth database and a fifth database;

[0013] The first database is used to construct a special vocabulary;

[0014] The second database includes a variety of gene symbols;

[0015] The third database includes a variety of gene-related literatures;

[0016] The fourth database includes relevant description information of diseases;

[0017] The fifth database includes the association between genes and phenotypes.

[0018] In an embodiment of the present application, the data integration module retrieves literatures in the third database according to the gene symbols in the second database, adds the relevant literatures to the special vocabulary of the first database, and forms a gene-disease-phenotype association form;

[0019] The data integration module retrieves the relevant description information of diseases in the fourth database according to the gene symbols in the second database, adds the relevant description information of the diseases to the special vocabulary of the first database, and forms a gene-disease-phenotype association form;

[0020] The data integration module supplements the association between genes and phenotypes in the fifth database to the special vocabulary of the first database, and forms a gene-disease-phenotype association form to obtain a multimodal database, and the multimodal database includes a special vocabulary in the form of a gene-disease-phenotype association.

[0021] In an embodiment of the present application, the knowledge graph construction module is used to construct a knowledge graph according to the data in the multimodal database, including:

[0022] The knowledge graph construction module constructs a knowledge graph according to a special vocabulary in the form of gene-disease-phenotype association in the multi-modal database.

[0023] In one embodiment of the present application, a voting mechanism is used to handle contradictions in multi-source data in the knowledge graph;

[0024] An algorithm is used to check the logical consistency of gene-phenotype-disease triples in the knowledge graph.

[0025] In one embodiment of the present application, the neural network module is used to process the preprocessed data according to the knowledge graph and the multi-modal database to obtain a pathogenic gene prediction result, including:

[0026] The knowledge graph and the multi-modal database are used to train the neural network module, and the trained neural network module processes the preprocessed data to obtain a pathogenic gene prediction result;

[0027] A dynamic weight allocation mechanism is used to optimize the neural network module to improve the prediction accuracy.

[0028] In one embodiment of the present application, the fast update module is used to update the multi-modal database, the knowledge graph and the neural network module, including:

[0029] The fast update module uses a third database to obtain the latest literature, adds the relevant literature to the special vocabulary of the first database, and forms a gene-disease-phenotype association form to update the multi-modal database;

[0030] The knowledge graph is updated according to the updated multi-modal database;

[0031] The neural network module is updated according to the updated knowledge graph and multi-modal database.

[0032] In one embodiment of the present application, the system further includes a data preprocessing module, and the data preprocessing module preprocesses the input data to obtain preprocessed data:

[0033] The preprocessing module checks the input data to obtain variant data;

[0034] A tool is used to perform variant annotation on the variant data;

[0035] According to the variant annotation, common variant data is removed and rare variant data is retained to obtain preprocessed data.

[0036] In one embodiment of the present application, the step of removing common variant data and retaining rare variant data according to the variant annotation to obtain preprocessed data includes:

[0037] The variant annotation of variant data includes population frequency. Variant data with a population frequency greater than the specified frequency is common variant data, and variant data with a population frequency not greater than the specified frequency is rare variant data.

[0038] According to another aspect of the present application, there is provided a storage medium on which a computer program is stored. When the computer program runs, it executes the above-mentioned gene prediction system.

[0039] The gene prediction system based on phenotype-gene association of the present invention includes a data integration module, a knowledge graph construction module, a neural network module, and a rapid update module. In this application, the data integration module integrates multiple medical databases to obtain a multi-modal database. The data in the multi-modal database is in the form of gene-disease-phenotype association. The neural network module processes the preprocessed data according to the knowledge graph and the multi-modal database to obtain the pathogenic gene prediction result. By associating phenotypes with genes, the accuracy of gene prediction is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] By describing the embodiments of the present application in more detail with reference to the accompanying drawings, the above and other objects, features, and advantages of the present application will become more apparent. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0041] Figure 1 The system block diagram of the gene prediction system based on phenotype-gene association according to the embodiment of the present application is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] In order to make the purpose, technical solutions, and advantages of the present application more obvious, the exemplary embodiments according to the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein. Based on the embodiments of the present application described in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0043] First, refer to Figure 1 to describe the gene prediction system 100 based on phenotype-gene association for implementing the embodiments of the present invention. Figure 1The system block diagram of a gene prediction system 100 based on phenotype-gene association according to an embodiment of the present application is shown. The gene prediction system 100 based on phenotype-gene association includes a data integration module 110, a knowledge graph construction module 120, a neural network module 130, and a rapid update module 140.

[0044] The data integration module 110 is used to integrate multiple medical databases to obtain a multi-modal database, and the data of the multi-modal database is in the form of gene-disease-phenotype association. The knowledge graph construction module 120 is used to construct a knowledge graph according to the data of the multi-modal database. The neural network module 130 is used to process the preprocessed data according to the knowledge graph and the multi-modal database to obtain a pathogenic gene prediction result. The rapid update module 140 is used to update the multi-modal database, the knowledge graph, and the neural network module 130.

[0045] In one embodiment, the data integration module 110 is used to integrate multiple medical databases to obtain a multi-modal database, including: the multiple medical databases include a first database, a second database, a third database, a fourth database, and a fifth database; the first database is used to construct a special vocabulary; the second database includes various gene symbols; the third database includes various literatures related to genes; the fourth database includes relevant description information of diseases; the fifth database includes the association between genes and phenotypes.

[0046] The first database can be the HPO (Human Phenotype Ontology) database. HPO is a medical phenotype ontology library based on standardized terms, founded by the Charité University Hospital. It contains more than 13,000 terms and 156,000 annotations for describing the clinical characteristics of genetic diseases. HPO terms are mainly formed in the form of a directed acyclic graph (DAG) and share information with other databases such as OMIM and MONDO. Establish a complete HPO term tree structure database from the HPO database, and perform manual proofreading and translation, including term definitions, synonyms (such as "fever" and "temperature elevation"), and hierarchical relationships. Then process the medical record corpus: collect de-identified clinical texts (outpatient records, discharge summaries, etc.), annotate the entity boundaries and corresponding HPO terms, and construct a special vocabulary for the medical field (including disease aliases, dialect expressions, etc.).

[0047] A second database might be the Human Genome Nomenclature Committee (HGNC), a nonprofit organization funded by HUGO (Human Genome Organization) that is responsible for naming human genes and providing authoritative gene symbols. HGNC manages a database of over 16,000 records, allowing users to search for gene information by symbol or name, including gene symbol, chromosome location, OMIM number, and PubMed ID.

[0048] The third database could be PubMed, an internet-based biomedical literature database developed by the National Center for Biotechnology Information (NCBI), part of the U.S. National Library of Medicine (NLM). PubMed indexes over 5,600 biomedical journals from over 70 countries and regions worldwide. These journals cover a wide range of fields, including medicine, biology, pharmacy, nursing, and dentistry.

[0049] The fourth database may be the OMIM (Online Mendelian Inheritance in Man) database. OMIM, the Online Mendelian Inheritance in Man, is an authoritative comprehensive database on human genes and genetic disorders.

[0050] The fifth database may be the MONDO (Monarch Disease Ontology) database, which is an ontology database for representing human diseases and related concepts, and combines cross-species gene, phenotype and disease data.

[0051] In one embodiment, the data integration module 110 searches for documents in the third database based on the gene symbols in the second database, adds the relevant documents to the special vocabulary of the first database, and forms a gene-disease-phenotype association form. The data integration module 110 searches for relevant descriptive information of the disease in the fourth database based on the gene symbols in the second database, adds the relevant descriptive information of the disease to the special vocabulary of the first database, and forms a gene-disease-phenotype association form. The data integration module 110 adds the association between genes and phenotypes in the fifth database to the special vocabulary of the first database, and forms a gene-disease-phenotype association form, thereby obtaining a multimodal database, which includes a special vocabulary of the gene-disease-phenotype association form.

[0052] Exemplarily, according to gene names, former names, etc. in the second database, retrieve literature related to human genes in the third database. After downloading the abstract part, use it to analyze whether it is relevant to the dedicated vocabulary. The specific method can be as follows: First, encode all entries in the dedicated vocabulary of the first database into vectors and store them as a benchmark vector library; at the same time, segment the literature abstract text into short sentences or encode the whole text into vectors. By calculating the cosine similarity between the abstract vector and the vocabulary vector, screen out the entries in the literature with a similarity higher than 0.8, and comprehensively judge the relevance by combining frequencies with exponential weighting. Select the literature with a relevance greater than 2 and incorporate it into the dedicated vocabulary of the first database and analyze the full text to form a gene-disease-phenotype association form. Score according to the SCI division of the literature, where the first division is scored 8, the second division is scored 4, the third division is scored 2, the fourth division is scored 1, and those not included in the division are scored 0.5.

[0053] Exemplarily, according to gene names, former names, etc. in the second database, parse the relevant description information of diseases in the fourth database. Since the disease relevance in the fourth database is relatively clear, direct association can be carried out. The specific parsing method can be as follows: First, encode all entries in the dedicated vocabulary of the first database into vectors and store them as a benchmark vector library; at the same time, segment the text of the disease-related description into short sentences or encode the whole text into vectors. By calculating the cosine similarity between the vector of the disease-related description information and the vocabulary vector, screen out the entries with a similarity higher than 0.8, and comprehensively judge the relevance by combining frequencies with exponential weighting to form a gene-disease-phenotype association form. The overall score for the relevant phenotype is 4.

[0054] Exemplarily, the fifth database mainly sorts out gene-related phenotypes reported in humans. Use the model to perform entity linking on the text description to supplement the missing gene-phenotype associations. Here, the number information, etc. in the fifth database can be linked to the dedicated vocabulary of the first database. Since its data source is mainly cross-species, the overall score for its relevant phenotypes is 0.25.

[0055] In one embodiment, the knowledge graph construction module 120 is used to construct a knowledge graph according to the data in the multi-modal database, including: the knowledge graph construction module 120 constructs a knowledge graph according to the dedicated vocabulary of the gene-disease-phenotype association form in the multi-modal database. Finally, a multi-modal knowledge graph containing 89,706 nodes (20,000 genes, 23,000 phenotypes, 11,000 diseases, 4,000 other entities) and 895,410 edges is generated.

[0056] In one embodiment, a voting mechanism is used to handle contradictions in multi-source data in the knowledge graph; an algorithm is used to verify the logical consistency of gene-phenotype-disease triples in the knowledge graph. To ensure the quality of the knowledge graph, a voting mechanism is adopted to handle contradictions in multi-source data (such as the pathogenicity annotation conflicts between Pubmed and OMIM). The logical consistency of gene-phenotype-disease triples is verified based on the Path Ranking Algorithm (PRA). Information related to OMIM is preferably adopted, followed by information related to the Pubmed database, and finally information related to the MONDO database.

[0057] In one embodiment, the neural network module 130 is configured to process preprocessed data based on the knowledge graph and the multi-modal database to obtain a pathogenic gene prediction result, including: training the neural network module 130 using the knowledge graph and the multi-modal database, and after training, the neural network module 130 processes the preprocessed data to obtain a pathogenic gene prediction result; optimizing the neural network module 130 using a dynamic weight allocation mechanism to improve the prediction accuracy.

[0058] Through message passing, the neural network module 130 can better capture complex structures and semantic information in graph data, thus achieving better performance in graph-related tasks. The message passing formula is:

[0059]

[0060] v is a node, is a positive integer, l is the number of layers, is a positive integer, u is a neighbor node, σ is an activation function, N v is the set of neighbor nodes of node v, a vu is the attention coefficient, W (l) is the weight matrix of, is the feature of neighbor node u of node v, is the new feature of node v.

[0061] According to the current known database situation, a simulated dataset of 50,000 patients is used for training. Finally, by introducing a dynamic weight allocation mechanism, the scores of the nodes or edges output by the neural network module 130 are optimized.

[0062] Let node v i have an original attention weight of a ij , and the dynamic weight correction formula is:

[0063] β i =σ(W d ·[Degree(v i )‖Sim(v i ,N(v i ))])

[0064] Degree(vi ) is the degree centrality of node v i , σ is the activation function, and W d is the learnable parameter matrix. N(v i ) is the set of neighbor nodes of node v i . Sim(v i , N(v i )) is the mean of the cosine similarities of the features between the node and its neighbor nodes. β i is the dynamic weight.

[0065]

[0066] a′ ij is the updated attention weight. γ is a scalar parameter used to control the weight of the similarity term. Τ is a parameter for adjusting the similarity magnitude. h i is the feature vector at position i, and h j is the feature vector at position j, which is used to measure the correlation between the two positions.

[0067] mainly enhances the attention weight of key nodes through the dynamic weight β i and introduces a Gaussian similarity term to alleviate local overfitting. The main purpose is to screen hub genes among them and use neighborhood similarity to filter out noisy edges.

[0068] In one embodiment, the fast update module 140 is used to update the multi-modal database, knowledge graph, and neural network module 130, including: the fast update module 140 uses the third database to obtain the latest literature, adds the relevant literature to the special vocabulary of the first database, and forms the gene-disease-phenotype association form to update the multi-modal database; updates the knowledge graph according to the updated multi-modal database; updates the neural network module 130 according to the updated knowledge graph and multi-modal database. By using the dynamic graph neural network, only local parameter updates are performed on the newly added subgraphs. In the graph convolutional layer, only feature aggregation is performed on the new subgraph and its neighborhood nodes. Among them, Anew is the updated adjacency matrix, and W is the parameter to be optimized

[0069]

[0070] is the node feature matrix of the l-th layer, is the adjacency matrix with self-loops, which is used to define the node connection relationship and support self-information aggregation. is the processed degree matrix, which is used to balance the influence of node degrees. W (l) is the trainable matrix, and σ is the activation function, is the node feature matrix of the (l + 1)-th layer.

[0071] In one embodiment, the system further includes a data preprocessing module, which preprocesses the input data to obtain preprocessed data: the preprocessing module checks the input data to obtain variant data; uses a tool to perform variant annotation on the variant data; and removes common variant data based on the variant annotation and retains rare variant data to obtain preprocessed data.

[0072] Data from 500 samples from a known laboratory dataset were processed and analyzed. The data included raw fastq data and basic sample information: sex, age, and HPO phenotype. Sequencing data quality was assessed using the FastQC tool, which checked sequence length, base quality (Q30), and GC content, removing low-quality or contaminated data. FastQC is a tool for quality control of high-throughput sequencing data. It supports multiple file formats (such as fastq, gzip-compressed fastq, SAM, and BAM), assesses the quality of raw sequencing reads (fastq format), and generates detailed sequence quality reports. Sequencing data were aligned to the reference genome (hg38) using the BWA tool, followed by Picard tagging of PCR duplicates to reduce false positives. BWA (Burrows-Wheeler Aligner) is an efficient tool for DNA sequence alignment, primarily used to map DNA sequences to large reference genomes, such as the human genome. Picard is a command-line tool for processing high-throughput sequencing data, commonly used for operations such as removing duplicate sequences, building indexes, and creating sequence dictionaries. The GATKHaplotypeCaller tool was used to identify single nucleotide variants (SNVs) and small insertions / deletions (Indels), generating VCF format variants. The GATKHaplotypeCaller is a tool for detecting single nucleotide variants (SNVs) and insertions / deletions (Indels).

[0073] In one embodiment, based on the variation annotation, common variation data is removed and rare variation data is retained to obtain preprocessed data, including: the variation annotation of the variation data includes population frequency, variation data with a population frequency greater than a specified frequency is common variation data, and variation data with a population frequency not greater than the specified frequency is rare variation data.

[0074] The Ensembl VEP tool was used to perform functional annotation of the mutations, including gene location (such as exons, introns), amino acid changes (missense, nonsense, frameshift, etc.), population frequencies (gnomAD, ExAC), conservation scores (REVEL), etc. Variants with relatively high population frequencies (frequency > 0.5% in gnomAD) were excluded, and rare variants were retained. Ensembl VEP (Ensembl Variant Effect Predictor) is a powerful tool for predicting the functional effects of genomic variants.

[0075] Basic sample information of the input samples: gender, age, HPO phenotypes. The filtered genes and mutations were predicted and ranked, and the system output was the ranking of the patient's genes, with the genes ranked higher being the pathogenic genes with a higher pathogenicity rate. Among 83.6% of the samples, the pathogenic genes were ranked in the top 10; among 90.5% of the samples, the pathogenic genes were ranked in the top 20; among 92.7% of the samples, the pathogenic genes were ranked in the top 50. The average analysis time for a single sample was 6.7 minutes, which was much lower than the manual analysis time of 41.34 minutes.

[0076] The laboratory datasets were analyzed using the gene prediction system of the present application, AMELIE, Exomiser, and PhenIX software respectively, and the comparison of the rankings of the pathogenic genes is shown in Table 1 below:

[0077] Table 1

[0078]

[0079]

[0080] It can be seen that the gene prediction system of the present application can rank more pathogenic genes in the top 50, obtaining the ranking of the genes most likely to cause disease.

[0081] The gene prediction system based on phenotype-gene association of the present invention includes a data integration module, a knowledge graph construction module, a neural network module, and a rapid update module. The present application integrates multiple medical databases through the data integration module to obtain a multi-modal database. The data in the multi-modal database is in the form of gene-disease-phenotype association. The neural network module processes the preprocessed data according to the knowledge graph and the multi-modal database to obtain the pathogenic gene prediction result. By associating phenotypes with genes, the accuracy of gene prediction is improved.

[0082] In addition, the present application also provides a storage medium, on which a computer program is stored. When the computer program is run by a processor, the processor is caused to execute the gene prediction system based on phenotype-gene association according to the embodiments of the present application described above. The storage medium may include, for example, a memory card of a smart phone, a storage component of a tablet computer, a hard disk of a personal computer, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disc read-only memory (CD-ROM), a USB memory, or any combination of the above storage media. The computer-readable storage medium may be any combination of one or more computer-readable storage media.

[0083] Although example embodiments have been described herein with reference to the drawings, it should be understood that the above example embodiments are merely exemplary and are not intended to limit the scope of the present application thereto. Those of ordinary skill in the art can make various changes and modifications therein without departing from the scope and spirit of the present application. All such changes and modifications are intended to be included within the scope of the present application as claimed in the appended claims.

[0084] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present application.

[0085] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0086] In the specification provided herein, a large number of specific details are set forth. However, it can be understood that the embodiments of the present application can be practiced without these specific details. In some instances, well-known methods, structures, and technologies have not been shown in detail so as not to obscure the understanding of this specification.

[0087] Similarly, it should be understood that, for the sake of streamlining this application and assisting in understanding one or more of the various inventive aspects, in the description of the exemplary embodiments of this application, the various features of this application are sometimes grouped together into a single embodiment, figure, or description thereof. However, the methods of this application should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected by the corresponding claims, the inventive point lies in that the corresponding technical problems can be solved with features less than all the features of a single disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into the detailed description, where each claim itself serves as a separate embodiment of this application.

[0088] Those skilled in the art will appreciate that, except where features are mutually exclusive, any combination can be used to combine all the features disclosed in this specification (including the accompanying claims, abstract, and drawings), as well as all the processes or units of any method or apparatus so disclosed. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) can be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0089] In addition, those skilled in the art can understand that, although some of the embodiments described herein include certain features included in other embodiments rather than other features, the combination of features of different embodiments means that it is within the scope of this application and forms different embodiments. For example, in the claims, any one of the claimed embodiments can be used in any combination.

[0090] The various component embodiments of this application can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some of the modules according to the embodiments of this application. This application can also be implemented as a program (e.g., a computer program and a computer program product) for performing part or all of the methods described herein. Such a program implementing this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or in any other form.

[0091] It should be noted that the above embodiments are illustrative of the present application rather than restrictive thereof, and those skilled in the art can design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in the claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present application can be implemented by means of hardware including several different elements and by means of a suitably programmed computer. The use of the words first, second, and third, etc. does not denote any order. These words can be interpreted as names.

[0092] As described above, the specific embodiments or descriptions of the specific embodiments of the present application are only illustrative, and the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should be covered by the protection scope of the present application. The protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A gene prediction system based on phenotype-gene association, characterized in that, It includes a data integration module, a knowledge graph construction module, a neural network module, and a rapid update module: The data integration module is used to integrate multiple medical databases to obtain a multi-modal database, and the data in the multi-modal database is in the form of gene-disease-phenotype association; The knowledge graph construction module is used to construct a knowledge graph according to the data in the multi-modal database; The neural network module is used to process the preprocessed data based on the knowledge graph and the multi-modal database to obtain a pathogenic gene prediction result; The rapid update module is used to update the multi-modal database, the knowledge graph, and the neural network module.

2. The gene prediction system based on phenotype-gene association according to claim 1, wherein The data integration module is used to integrate multiple medical databases to obtain a multi-modal database, including: The multiple medical databases include a first database, a second database, a third database, a fourth database, and a fifth database; The first database is used to construct a special vocabulary; The second database includes a variety of gene symbols; The third database includes a variety of literatures related to genes; The fourth database includes relevant description information of diseases; The fifth database includes the association between genes and phenotypes.

3. The gene prediction system based on phenotype-gene association according to claim 2, wherein The data integration module retrieves literatures in the third database according to the gene symbols in the second database, adds the relevant literatures to the special vocabulary of the first database, and forms the gene-disease-phenotype association form; The data integration module retrieves the relevant description information of diseases in the fourth database according to the gene symbols in the second database, adds the relevant description information of the diseases to the special vocabulary of the first database, and forms the gene-disease-phenotype association form; The data integration module supplements the association between genes and phenotypes in the fifth database to the special vocabulary of the first database, and forms the gene-disease-phenotype association form to obtain a multi-modal database, and the multi-modal database includes a special vocabulary in the form of gene-disease-phenotype association.

4. The gene prediction system based on phenotype-gene association according to claim 3, wherein The knowledge graph construction module is used to construct a knowledge graph according to the data in the multi-modal database, including: The knowledge graph construction module constructs a knowledge graph according to the special vocabulary in the form of gene-disease-phenotype association in the multi-modal database.

5. The gene prediction system based on phenotype-gene association according to claim 4, wherein Use a voting mechanism to handle the contradictions of multi-source data in the knowledge graph; Use an algorithm to check the logical consistency of the gene-phenotype-disease triples in the knowledge graph.

6. The gene prediction system based on phenotype-gene association according to claim 1, characterized in that, The neural network module is used to process the preprocessed data based on the knowledge graph and the multi-modal database to obtain a pathogenic gene prediction result, including: Use the knowledge graph and the multi-modal database to train the neural network module, and the trained neural network module processes the preprocessed data to obtain a pathogenic gene prediction result; Use a dynamic weight allocation mechanism to optimize the neural network module to improve the prediction accuracy.

7. The gene prediction system based on phenotype-gene association according to claim 3, wherein The rapid update module is used to update the multi-modal database, the knowledge graph, and the neural network module, including: The fast update module uses the third database to obtain the latest literature, adds the relevant literature to the special vocabulary of the first database, and forms a gene-disease-phenotype association form to update the multi-modal database; Update the knowledge graph according to the updated multi-modal database; Update the neural network module according to the updated knowledge graph and multi-modal database.

8. The gene prediction system based on phenotype-gene association according to claim 1, wherein The system further includes a data preprocessing module, and the data preprocessing module preprocesses the input data to obtain preprocessed data: The preprocessing module checks the input data to obtain variant data; Use tools to perform variant annotation on the variant data; According to the variant annotation, remove the common variant data and retain the rare variant data to obtain the preprocessed data.

9. The gene prediction system based on phenotype-gene association according to claim 8, wherein The step of removing the common variant data and retaining the rare variant data according to the variant annotation to obtain the preprocessed data includes: The variant annotation of the variant data includes the population frequency. The variant data with a population frequency greater than the specified frequency is common variant data, and the variant data with a population frequency not greater than the specified frequency is rare variant data.

10. A storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program runs, it executes the gene prediction system according to any one of claims 1-9.

Images