WO2025065483A1 - Genetic locus mutation prognostic risk assessment method, electronic device and storage medium - Google Patents

Abstract

Provided in the embodiments of the present disclosure are a genetic locus mutation prognostic risk assessment method, an electronic device and a storage medium. The method comprises: acquiring an initial data set and a control data set, the initial data set being obtained on the basis of at least one genetic locus of a target gene being mutated, and the control data set being obtained on the basis of none of the gene loci of the target gene being mutated; preprocessing the initial data set to obtain a target gene mutation data set; using the proportional hazards model to determine a prognostic risk value of each gene locus in the target gene mutation data set; according to the prognostic risk values, selecting a target gene locus; and using an ensemble classification model to obtain a prognostic risk type result of the target gene locus.

Description

Prognostic risk assessment method for gene site mutation, electronic device and storage medium Technical Field

The present application relates to the fields of biomedical technology and artificial intelligence technology, and in particular to a prognostic risk assessment method for gene site mutations, an electronic device, and a storage medium.

Background Art

Prognostic risk assessment of the prognostic risk caused by gene mutation can provide a reference for doctors to formulate clinical plans, but the prognostic risk assessment schemes in related technologies cannot accurately assess the prognostic risk caused by gene mutation.

Summary of the invention

According to a first aspect, the present disclosure provides a method for assessing the prognostic risk of gene site mutations, comprising: obtaining an initial data set and a control data set, wherein the initial data set is obtained based on a mutation in at least one gene site of a target gene, and the control data set is obtained based on the fact that none of the gene sites of the target gene have mutated; preprocessing the initial data set to obtain a target gene mutation data set; determining the prognostic risk value of each gene site in the target gene mutation data set using a proportional risk regression model; selecting a target gene site based on the prognostic risk value; and obtaining the prognostic risk type result of the target gene site using an integrated classification model.

For example, the above-mentioned initial data set includes multiple gene mutation data; the above-mentioned preprocessing of the above-mentioned initial data set to obtain the target gene mutation data set includes: based on the preset gene annotation requirements, data combing of the multiple initial gene mutation data to obtain multiple gene mutation data to be processed; using the preset gene annotation statement, annotating the multiple gene mutation data to be processed respectively to obtain multiple target gene mutation data, wherein each of the above-mentioned target gene mutation data includes multiple transcript information sets and multiple amino acid information corresponding to the above-mentioned gene site where the mutation occurs; and obtaining the above-mentioned target gene mutation data set based on the multiple above-mentioned target gene mutation data.

For example, obtaining the target gene mutation data set according to the plurality of target gene mutation data includes: acquiring target transcript information and target amino acid information associated with the target gene from a transcript database; filtering the plurality of transcript information and the plurality of amino acid information included in each of the target gene mutation data according to the target transcript information and the target amino acid information, so that only the target transcript information and the target amino acid information are retained among the plurality of transcript information and the plurality of amino acid information of each of the target gene mutation data; and And generating the target gene mutation data set according to the plurality of target gene mutation data retaining the target transcript information and the target amino acid information.

For example, the above-mentioned use of the proportional risk regression model to determine the prognostic risk value of each gene site in the above-mentioned target gene mutation data set includes: based on the above-mentioned target gene mutation data set, determining the sub-mutation data corresponding to each of the above-mentioned gene sites; using the proportional risk regression model to perform a prognostic risk quantitative analysis between each of the above-mentioned sub-mutation data and the above-mentioned control data set to obtain the prognostic risk value of the above-mentioned gene site.

For example, the target gene is linked to the target gene mutation data set, the target gene includes m amino acid sites, each of the amino acid sites corresponds to at least one gene site where a mutation occurs, and m is a positive integer; based on the target gene mutation data set, determining the sub-mutation data corresponding to each of the gene sites includes: sliding a sliding window on the target gene, wherein the center point of the sliding window is any one of the m amino acid sites, the size of the sliding window along the extension direction of the target gene is associated with the distance between k adjacent amino acid sites along the extension direction of the target gene, and k is a positive integer; when the sliding window slides to any one of the amino acid sites, based on the size of the sliding window along the extension direction of the target gene, k amino acid sites are selected around the amino acid site as the center to obtain the target amino acid site; and the gene mutation data in the target gene mutation data set linked to the target amino acid site is used as the sub-mutation data.

For example, the prognostic risk value of each of the above-mentioned gene sites is characterized by the prognostic risk value of the amino acid site associated with the above-mentioned gene site, and the above-mentioned sub-mutation data is associated with the above-mentioned amino acid site; the above-mentioned use of the proportional risk regression model to perform a quantitative prognostic risk analysis between each of the above-mentioned sub-mutation data and the above-mentioned control data set to obtain the prognostic risk value of the above-mentioned gene site includes: inputting the above-mentioned sub-mutation data into the prognostic risk function to obtain the prognostic risk function value corresponding to each of the above-mentioned amino acid sites; and taking the ratio between the prognostic risk function value of each of the above-mentioned amino acid sites and the benchmark prognostic risk function value as the prognostic risk value of each of the above-mentioned amino acid sites, wherein the above-mentioned benchmark prognostic risk function value is obtained by inputting the above-mentioned control data set into the above-mentioned prognostic risk function.

For example, the method further includes: drawing a mutation prognosis risk value scatter plot based on the prognosis risk value of each of the amino acid sites; and displaying the mutation prognosis risk value scatter plot using a visualization component.

For example, the above-mentioned selecting the target gene site according to the above-mentioned prognostic risk value includes: according to the prognostic risk value of each of the above-mentioned amino acid sites, selecting the target amino acid site from the amino acid sites associated with the above-mentioned gene site, wherein the above-mentioned target amino acid site includes the above-mentioned target gene site.

For example, the above-mentioned selecting a target amino acid site from the amino acid sites associated with the above-mentioned gene site based on the prognostic risk value of each of the above-mentioned amino acid sites includes: calculating the credibility associated with the prognostic risk value of each of the above-mentioned amino acid sites; comparing the credibility associated with the prognostic risk value of each of the above-mentioned amino acid sites with a credibility threshold to obtain a prognostic risk value less than the above-mentioned credibility threshold; determining the amino acid site associated with the prognostic risk value less than the above-mentioned credibility threshold from the amino acid sites associated with the above-mentioned gene site to obtain the above-mentioned target amino acid site.

For example, the above-mentioned prognostic risk type result of the above-mentioned target gene site obtained by using the integrated classification model includes: for the prognostic risk caused by the mutation of the above-mentioned target amino acid site, the above-mentioned integrated classification model is used to perform a qualitative analysis of the prognostic risk type to obtain the above-mentioned prognostic risk type result.

For example, the above-mentioned prognostic risk caused by the mutation of the above-mentioned target amino acid site is qualitatively analyzed by using the above-mentioned integrated classification model to obtain the above-mentioned prognostic risk type result, which includes: based on the above-mentioned target amino acid site, extracting the gene mutation data associated with the mutation of the above-mentioned target amino acid site from the above-mentioned target gene mutation data set to obtain the mutation data of the prognostic risk type to be analyzed; based on the mutation data of the above-mentioned prognostic risk type to be analyzed, using the above-mentioned integrated classification model to predict the degree of match between the prognostic risk caused by the mutation of the above-mentioned target amino acid site and each of the N prognostic risk types, to obtain N predicted matching values corresponding to the N above-mentioned prognostic risk types, N is a positive integer, and the prognostic risk degrees between the N prognostic risk types are different from each other; and determining the above-mentioned prognostic risk type result according to the N predicted matching values.

For example, based on the mutation data of the above-mentioned prognostic risk type to be analyzed, the above-mentioned integrated classification model is used to predict the degree of match between the prognostic risk caused by the mutation at the above-mentioned target amino acid site and each of the N prognostic risk types, and N predicted matching values corresponding to the N above-mentioned prognostic risk types are obtained, including: for each of the N prognostic risk types: using the integrated classification model to analyze the prognostic risk caused by the mutation at the above-mentioned target amino acid site and the degree of match between the above-mentioned prognostic risk type, to obtain multiple prediction values, wherein the above-mentioned integrated classification model includes multiple classifiers, each classifier can obtain one of the above-mentioned prediction values; using the gradient boosting classifier to process the multiple above-mentioned prediction values to obtain the above-mentioned prediction matching value.

For example, determining the above-mentioned prognostic risk type result based on N predicted matching values includes: selecting the highest predicted matching value from the N predicted matching values by performing numerical comparison between the above-mentioned N predicted matching values; and taking the prognostic risk type corresponding to the above-mentioned highest predicted matching value as the above-mentioned prognostic risk type result.

For example, the method further includes: acquiring the prognostic risk type result; and pushing the prognostic risk type result to the target object, so that the target object generates a clinical decision according to the prognostic risk type result.

For example, the above-mentioned integrated classification model is trained in the following manner: obtaining a sample data set, wherein the above-mentioned sample data set includes first prognostic risk level data and second prognostic risk level data, and the above-mentioned first prognostic risk level is higher than the above-mentioned second prognostic risk level; randomly splitting the above-mentioned sample data set into an initial training sample set and a verification sample set, wherein the ratio of the number of data in the above-mentioned initial training sample set to the number of data in the above-mentioned test sample data set is a preset ratio; oversampling the above-mentioned initial training sample set so that the number of the above-mentioned first prognostic risk level data in the above-mentioned initial training sample set is the same as the number of the above-mentioned second prognostic risk level data, and obtaining a target training sample set; cross-validation training is performed on the initial integrated classification model using the above-mentioned target training sample set to obtain an intermediate integrated classification model; the above-mentioned intermediate integrated classification model that meets the preset training end condition is used as the above-mentioned integrated classification model, wherein the above-mentioned preset training end condition includes that the number of training times reaches a preset training time threshold.

For example, the cross-validation training of the initial integrated classification model using the target training sample set includes: randomly dividing the target training sample set into Q training sample subsets, where Q is a positive integer; repeatedly performing the following operations until the preset training end condition is met: selecting Q-1 of the Q training sample subsets to train the initial integrated classification model to obtain an intermediate integrated classification model for the next round of training, wherein the Q-1 training sample subsets selected are different from each training session to each training session; and using the remaining 1 of the Q training sample subsets to test the intermediate integrated classification model for the next round of training, wherein the training sample subsets used are different from each test to each test.

For example, the sample data set includes multiple sample data, and the types of the sample data include at least one of the following: data on the impact of the mutated target amino acid site on protein function, data on the clinical impact of the mutated target amino acid site, and position data of the mutated target amino acid site.

For example, the above classifier includes at least one of the following: a random forest classifier, a Gini index random tree classifier, an entropy random tree classifier, and a gradient boosting classifier.

According to the second aspect, the present disclosure also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned prognostic risk assessment method for gene site mutations.

According to the third aspect, the present disclosure also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above-mentioned prognostic risk assessment method for gene site mutations.

According to a fourth aspect, the present disclosure further provides a computer program product, including a computer program, which implements the above-mentioned prognostic risk assessment method for gene site mutation when executed by a processor.

BRIEF DESCRIPTION OF THE DRAWINGS

The above and other features of the embodiments of the present disclosure will become more apparent through the following detailed description in conjunction with the accompanying drawings, in which:

FIG1 is a flow chart of a method for assessing the prognostic risk of gene site mutations according to an embodiment of the present disclosure;

FIG2 is a schematic diagram of an initial data set according to an embodiment of the present disclosure;

FIG3 is a schematic diagram of a control data set according to an embodiment of the present disclosure;

FIG4 is a schematic diagram of gene mutation data to be processed according to an embodiment of the present disclosure;

FIG. 5 is data after gene re-annotation according to an embodiment of the present disclosure

FIG6 is an example diagram of a target gene mutation dataset according to an embodiment of the present disclosure;

FIG7 is a schematic diagram of collecting data using a sliding window according to an embodiment of the present disclosure;

FIG8 is a scatter plot of mutation prognostic risk values according to an embodiment of the present disclosure;

FIG9 is a schematic diagram of an architecture for obtaining a predicted matching value using an integrated classification model according to an embodiment of the present disclosure;

FIG10A is a schematic diagram of data obtained by using some tools to predict amino acid site mutations according to an embodiment of the present disclosure;

FIG10B is a schematic diagram of data obtained by using some tools to predict amino acid site mutations according to another embodiment of the present disclosure;

FIG11A is a schematic diagram of clinical effects of amino acid site mutations according to an embodiment of the present disclosure;

FIG11B is a schematic diagram of clinical effects of amino acid site mutations according to another embodiment of the present disclosure;

FIG12A is a ROC curve of the integrated classification model of an embodiment of the present disclosure;

FIG12B is a ROC curve of the related art model;

FIG13 is a method for assessing the prognostic risk of gene site mutations according to another embodiment of the present disclosure; and

FIG. 14 is a block diagram of an electronic device suitable for implementing an image processing method according to an embodiment of the present disclosure.

In the drawings, the same or similar structures are marked with the same or similar reference numerals.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the described embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work belong to the protection of the present disclosure. The present invention relates to a device for carrying out the present invention and to a device for carrying out the present invention. It should be noted that throughout the accompanying drawings, the same elements are represented by the same or similar reference numerals. In the following description, some specific embodiments are used for descriptive purposes only and should not be understood as any limitation to the present disclosure, but are merely examples of embodiments of the present disclosure. Conventional structures or configurations will be omitted when they may cause confusion in the understanding of the present disclosure. It should be noted that the shapes and sizes of the components in the figures do not reflect the actual size and proportion, but only illustrate the contents of the embodiments of the present disclosure.

It should be noted that in the drawings, the size and relative size of the elements may be exaggerated for the purpose of clarity and/or description. Thus, the size and relative size of each element are not necessarily limited to the size and relative size shown in the drawings. In the specification and drawings, the same or similar reference numerals indicate the same or similar parts.

Unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should be the common meanings understood by those of ordinary skill in the art. The "first", "second" and similar words used in the present disclosure do not represent any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the element or object appearing before the word covers the element or object listed after the word and its equivalent, without excluding other elements or objects.

In this document, unless otherwise specified, directional terms such as "upper", "lower", "left", "right", "inner", "outer", etc. are used to indicate the orientation or positional relationship based on the drawings, and are only for the convenience of describing the present disclosure, and do not indicate or imply that the device, element or component referred to must have a specific orientation, be constructed or operate in a specific orientation. It should be understood that when the absolute position of the described object changes, the relative positional relationship they represent may also change accordingly. Therefore, these directional terms should not be understood as limiting the present disclosure.

In addition, in the description of the embodiments of the present disclosure, the term "connected" or "connected to" may refer to two components being directly connected, or may refer to two components being connected via one or more other components. In addition, the two components may be connected or coupled via a wired or wireless manner.

In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of the data involved (such as but not limited to user personal information), the process of obtaining the initial data set about the user, the process of obtaining the reference data set about the user, etc., all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

According to an example, the clinical significance of somatic mutations in cancer is explained through the following scheme: 5234 somatic mutations from the user's clinical report database are obtained as training sets and validation sets, and 6226 mutations are obtained as test sets through literature retrieval; using functional and clinical evidence, the semi-supervised generative adversarial network (SGAN) method is used to predict the carcinogenicity of mutations; somatic mutations are classified into 4 categories: strong clinical significance, potential clinical significance, and clinical significance. uncertain and benign/likely benign.

The above example only explains the clinical significance of somatic mutations for cancer. However, since clinical significance is not equivalent to prognostic risk, it is not possible to accurately judge whether somatic mutations are protective or harmful to prognostic risk based on clinical interpretation, which reduces the accuracy of assessing the prognostic risk of gene mutations. Prognosis refers to the prediction of the development process and final outcome of a disease after measures are taken.

In view of this, the embodiments of the present disclosure provide a method for assessing the prognostic risk of gene site mutations, which analyzes the impact of each gene site mutation on the prognostic risk of cancer based on the prognostic risk, and improves the accuracy of assessing the prognostic risk of gene mutations.

FIG1 is a flow chart of a method for assessing the prognostic risk of gene site mutations according to an embodiment of the present disclosure.

As shown in FIG. 1 , the method for assessing the prognostic risk of gene site mutation may include operations S110 to S150 .

In operation S110, an initial data set and a control data set are obtained, wherein the initial data set is obtained based on at least one gene site of the target gene being mutated, and the control data set is obtained based on no gene sites of the target gene being mutated.

In operation S120, the initial data set is preprocessed to obtain a target gene mutation data set.

In operation S130 , a proportional hazard regression model is used to determine the prognostic risk value of each gene site in the target gene mutation dataset.

In operation S140, a target gene locus is selected according to the prognostic risk value.

In operation S150, the prognostic risk type result of the target gene site is obtained using the integrated classification model.

According to an embodiment of the present disclosure, the database may include at least one of the following: the public ICGC (International Cancer Genome Consortium) database and the public MSK (Memorial Sloan Kettering Cancer Center) database.

According to the embodiments of the present disclosure, taking the target gene as the TP53 gene as an example, the TP53 gene may include 393 amino acid sites, each amino acid site corresponds to 3 bases, and a mutation in a gene site may refer to a mutation in a base, and the mutation in the gene site will cause the amino acid corresponding to the base to change, and the change in the amino acid will bring about prognostic risk. Therefore, it is necessary to perform a prognostic risk assessment on the prognostic risk caused by the mutation of each gene site on the gene, or on the mutation of each amino acid site.

According to the embodiments of the present disclosure, the initial data set and the control data set can be used to evaluate the prognostic risk value caused by each gene site mutation or each amino acid site mutation on the target gene (such as TP53). And for the prognostic risk value with low credibility, the clinical significance of the amino acid site mutation corresponding to the prognostic risk value can be predicted. Prognostic risk level. It is understood that the prognostic risk assessment method according to the embodiment of the present disclosure can be used to assess the prognostic risk brought about by mutations at any other gene site or any amino acid site.

FIG. 2 is a schematic diagram of an initial data set according to an embodiment of the present disclosure.

As shown in Figure 2, the initial data set may include multiple initial gene mutation data, and the initial gene mutation data is obtained based on the mutation of at least one gene site of the target gene. Specifically, the multiple initial gene mutation data may include clinical data and missense mutation data. Each row in Figure 2 can be regarded as the initial gene mutation data of a user. In each row, the clinical data may refer to the user's number, the user's survival status, the user's survival time, etc. The missense mutation data may include the mutant gene to be analyzed, the chromosome number, the starting position of the gene site, the ending position of the gene site, the reference gene, and the mutant gene.

Continuing with FIG. 2 , P- in the number column is used to represent the user ID of the initial data set, and n can represent the number of users, where n is a positive integer. The “+” in the chain column is used to represent the positive chain, which refers to the chain on DNA that carries the nucleotide sequence encoding the amino acid information of the protein. The positive chain is also called the coding chain, meaningful chain or positive chain (+ chain).

Regarding the reference gene and mutant gene in Figure 2, taking the first row in Figure 2 as an example, the reference gene at the 7577121 gene site should be G (guanine), but due to gene mutation, it mutated into the mutant gene A (adenine).

The units of measurement for survival time in Figure 2 can be years, months, days, hours, etc., and can be represented by the symbol t or time. The distribution of survival time is generally not a normal distribution. Survival time can be the time between the starting event and the ending event. The starting event can be diagnosis, medication, surgical resection, etc., and the ending event can be recovery, death, remission, recurrence, etc.

FIG. 3 is a schematic diagram of a comparison data set according to an embodiment of the present disclosure.

As shown in Figure 3, the control data set refers to the data obtained when there is no mutation in the gene locus on the target gene (such as TP53), which is used as the control data of the initial gene mutation data. The acquisition of the control data set can also be obtained from at least one of the public ICGC database and the public MSK database. Each row in Figure 3 can be regarded as the control data of a user. Specifically, the control data may include the user's number, the user's survival status, and the user's survival time. In the user's number column, D- is used for the user ID of the control data set.

According to an embodiment of the present disclosure, the process of preprocessing multiple initial gene mutation data may include data combing of the initial gene mutation data, gene re-annotation, and the like, and the target gene mutation data set obtained after preprocessing may be a data set including transcript and amino acid information.

According to the embodiments of the present disclosure, the target gene locus may be a reliable clinical For the gene loci of the prognostic risk result, for example, due to the small amount of sample data based on which the prognostic risk value is obtained and the low credibility, it is impossible to obtain a reliable clinical prognostic risk result based on the prognostic risk value. Such gene loci need to be further analyzed for prognostic risk qualitatively using an integrated classification model to obtain a reliable clinical prognostic risk result. Optionally, since changes in gene loci will lead to changes in amino acid loci, in the process of qualitative analysis of target gene loci, the target amino acid loci can be used instead of the analysis of the target gene loci.

According to an embodiment of the present disclosure, the process of selecting a target gene locus from a gene locus according to a prognostic risk value may include selecting a target gene locus from a gene locus according to conditions such as the amount of sample data based on which the prognostic risk value is obtained is less than a preset sample data amount threshold, and/or the credibility is less than a credibility threshold. Alternatively, a target amino acid locus including a target gene locus may also be selected from an amino acid locus according to conditions such as the amount of sample data based on which the amino acid prognostic risk value is obtained is less than a preset sample data amount threshold, and/or the credibility is less than a credibility threshold.

According to the embodiment of the present disclosure, by obtaining an initial data set and a control data set; preprocessing the initial data set to obtain a target gene mutation data set; using a proportional risk regression model to determine the prognostic risk value of each gene site in the target gene mutation data set; selecting a target gene site according to the prognostic risk value; using an integrated classification model to obtain the prognostic risk type result of the target gene site. Since the embodiment of the present disclosure can realize a quantitative prognostic risk assessment of the prognostic risk value of the mutation gene site, the mutation prognostic risk value of each gene site is obtained, and the target gene site with low credibility of the prognostic risk value and unknown clinical significance is qualitatively assessed to obtain the prognostic risk type of the target gene site. The prognostic risk assessment method for gene site mutation provided by the embodiment of the present disclosure can not only quantify the prognostic risk of gene mutation in a more fine-grained manner, but also qualitatively analyze the target gene site of unknown clinical significance, thereby at least partially overcoming the problem of inaccurate assessment of the prognostic risk of gene mutation in the related art, thereby improving the accuracy of the prognostic risk assessment of gene mutation, and providing more comprehensive reference information for the formulation of clinical plans for doctors.

According to an embodiment of the present disclosure, since different reference genomes are used to annotate mutant genes in the ICGC database or the MSK database, there may be multiple annotation results for synonymous mutations of the same gene. Therefore, it is necessary to use an annotation tool to re-annotate the initial gene mutation data shown in Figure 2, hereinafter referred to as gene re-annotation. Specifically, operation S120 may include the following operations: based on the preset gene annotation requirements, a plurality of initial gene mutation data are combed to obtain a plurality of gene mutation data to be processed; using a preset gene annotation statement, a plurality of gene mutation data to be processed are annotated and processed respectively to obtain a plurality of target gene mutation data, wherein each target gene mutation data includes a plurality of transcript information sets and a plurality of amino acid information corresponding to the gene site where the mutation occurs; and a target gene mutation data set is obtained based on a plurality of target gene mutation data.

FIG. 4 is a schematic diagram of gene mutation data to be processed according to an embodiment of the present disclosure; FIG. 5 is data after gene re-annotation according to an embodiment of the present disclosure.

As shown in FIG4 , the preset gene annotation requirements may refer to the need to retain the chromosome number column, gene site start position column, gene site end position column, reference gene column, and mutant gene column in the gene mutation data. After the initial gene mutation data is sorted according to the preset annotation requirements, the obtained multiple gene mutation data to be processed can be shown in FIG4 . In addition to the chromosome number column, gene site start position column, gene site end position column, reference gene column, and mutant gene column in FIG4 , multiple custom columns can also be set according to the implementation situation. FIG3 only takes custom column 1 as a chain, "1" in the chain column is used to represent the positive chain, and custom column 2 is a user number as an example.

According to the embodiments of the present disclosure, the annotation tool can be adaptively adjusted according to actual needs. The embodiments of the present disclosure are described by taking the Annovar annotation tool as an example. Multiple gene mutation data to be processed are input into the annotation tool, and the gene annotation statement annotate_variation.pl-geneanno-dbtype refGene-buildver hg19 example/ex1.avinput humandb/ is used to perform gene re-annotation on the multiple gene mutation data to be processed. Among them, -buildver hg19 can indicate that the reference genome uses the hg19 reference gene. After the multiple gene mutation data to be processed shown in Figure 4 are re-annotated using the above-mentioned gene annotation statement, the obtained data can be shown in Figure 5.

Continuing to refer to FIG5 , the gene mutation types shown in the second column and the transcript information and amino acid information shown in the third column are obtained after re-annotating the multiple gene mutation data to be processed shown in FIG4 . Among them, a transcript is one or more mature mRNAs (messenger Ribonucleic Acid, messenger RNA) that can encode proteins formed by transcription of a gene. In the second column, N can represent a nonsynonymous mutation type, S can represent a stoploss mutation type, and F can represent a frame shift mutation.

According to an embodiment of the present disclosure, since a gene generally has multiple transcripts, that is, the third column in Figure 5 will contain multiple transcript information and corresponding amino acid information, and each row in Figure 5 may also contain multiple transcript information and corresponding amino acid information. Exemplarily, in a complete transcript, the information included may be ENST00000642122.1_4: exon10: c.G1641A: p.M547I. Among them, p.M547I may be the amino acid information corresponding to the transcript. In one embodiment, the more information the transcript includes, the more comprehensive the gene mutation information covered will be. Based on this, the data after the gene re-annotation shown in Figure 5 can be filtered and screened for transcript information.

According to the embodiments of the present disclosure, by performing preprocessing such as data combing and gene re-annotation on the initial gene mutation data, the format of the initial gene mutation data can be made neat and unified, thereby improving the efficiency and accuracy of the prognostic risk assessment of gene mutations when evaluating the prognostic risk of gene mutations.

According to an embodiment of the present disclosure, obtaining a target gene mutation data set based on multiple target gene mutation data may include the following operations: obtaining target transcript information and target amino acid information associated with the target gene from a transcript database; filtering the multiple transcript information and multiple amino acid information included in each target gene mutation data based on the target transcript information and the target amino acid information, so that only the target transcript information and the target amino acid information are retained among the multiple transcript information and multiple amino acid information of each target gene mutation data; and generating a target gene mutation data set based on the multiple target gene mutation data that retain the target transcript information and the target amino acid information.

According to an embodiment of the present disclosure, the target transcript information may refer to the transcript information containing the most information obtained from a specified database. The target transcript information is generally the most classic transcript information, generally the most correct transcript information with a reliable source and verified by multiple parties. The target amino acid information may be the amino acid information in the target transcript information.

FIG. 6 is an example diagram of a target gene mutation dataset according to an embodiment of the present disclosure.

As shown in Figure 6, the target transcript information and target amino acid information are used to filter and screen each row of information in the third column shown in Figure 5, so that each row only retains the target transcript information and target amino acid information. After screening, the transcript information can be simplified, and only the reference amino acid information, the amino acid information after mutation, and the mutant amino acid position are extracted, and the target gene mutation data set obtained can be shown in Figure 6. Among them, the reference amino acid information and the amino acid information after mutation can refer to the amino acid information of the five columns; the mutant amino acid position can refer to the mutation position of the amino acid site in the sixth column. Exemplarily, taking the mutant amino acid p.Y279C as an example, it can be indicated that the reference amino acid tyrosine (Y) with the mutant amino acid position of No. 279 becomes the mutant amino acid cysteine (C). The MV of the consequence column in Figure 6 can represent a missense mutation (missense_variant).

According to the embodiments of the present disclosure, after simplifying the transcript information, extracting the reference amino acid information, the amino acid information after mutation and the position of the mutated amino acid, and before obtaining the target gene mutation data set, the data set can be preprocessed to fill in missing values and delete abnormal values. For example, the abnormal data in the survival status and survival time columns are processed, and the values with negative survival time are deleted; or garbled characters and characters that are not numbers are deleted. In the case of missing survival time, it is also possible to fill in the duration obtained according to the follow-up time or delete the user data corresponding to the missing value according to the survival status, etc. The specific situation can be adaptively adjusted according to actual needs.

According to the embodiments of the present disclosure, by using the target transcript information and the target amino acid information, the multiple transcript information and the multiple amino acid information included in each target gene mutation data are filtered. Since the target transcript information can be the most correct transcript information verified by multiple parties, only the information associated with the target transcript information and the target amino acid information can be retained after filtering the multiple transcript information and the multiple amino acid information, thereby ensuring the accuracy of the information. The simplicity and accuracy of information can improve the efficiency and accuracy of prognostic risk assessment of gene mutations.

According to the embodiments of the present disclosure, after obtaining the target gene mutation data set, a sliding window and a proportional-hazards model (also known as a Cox regression model) can be used to analyze the prognostic risk value after a single amino acid site mutation. It should be noted that the above-mentioned data preprocessing stage is to obtain and process data with the gene site as the minimum granularity, but the mutation of the gene site will cause the amino acid corresponding to the gene site to change. Therefore, in order to simplify the workload of the evaluation, improve the efficiency of the evaluation of the prognostic risk of the mutation, and be able to describe the prognostic risk value of the mutation more clearly and concisely, the embodiments of the present disclosure analyze the prognostic risk value of the mutation in units of amino acid sites.

According to an embodiment of the present disclosure, operation S130 may include the following operations: based on the target gene mutation data set, determining the sub-mutation data corresponding to each gene site; using a proportional risk regression model to perform a quantitative prognostic risk analysis between each sub-mutation data and the control data set to obtain the prognostic risk value of the gene site.

According to an embodiment of the present disclosure, the sub-mutation data may be data corresponding to each gene site where a gene mutation occurs. Since the mutation of a gene site affects the corresponding amino acid, the sub-mutation data may also be data corresponding to an amino acid site where a gene site mutates. The process of obtaining the sub-mutation data may be obtained by sliding a sliding window on the target gene chain. Specifically, based on the target gene mutation data set, determining the sub-mutation data corresponding to each gene site may include the following operations: sliding a sliding window on the target gene, wherein the center point of the sliding window is any one of the m amino acid sites, the size of the sliding window along the extension direction of the target gene, and the distance between the k adjacent amino acid sites along the extension direction of the target gene are associated, and k is a positive integer; when the sliding window slides to any one amino acid site, based on the size of the sliding window along the extension direction of the target gene, k amino acid sites are selected around the amino acid site as the center to obtain the target amino acid site; and the gene mutation data linked to the target amino acid site in the target gene mutation data set is used as the sub-mutation data.

According to an embodiment of the present disclosure, a target gene and a target gene mutation dataset can be linked, the target gene includes m amino acid sites, each amino acid site corresponds to at least one gene site where a mutation occurs, and m is a positive integer. For example, taking TP53 as an example, TP53 includes 393 amino acid sites, and m can be 393.

FIG. 7 is a schematic diagram of collecting data using a sliding window according to an embodiment of the present disclosure.

As shown in FIG. 7 , the size B of the sliding window 701 along the extension direction of the target gene can be the size including k adjacent amino acid sites, and FIG. 7 is shown by taking k equal to 3 as an example. By sliding the sliding window 701 on the target gene chain 702, the size B along the extension direction of the target gene with amino acid site i (0≤i≤392) as the center can be obtained. B is k, and the area enclosed is a sliding window of (ik/2, i+k/2). The mutation data within the area enclosed by the sliding window are collected to obtain sub-mutation data. Among them, k can be adaptively adjusted according to actual needs. When i=0, the area enclosed by the sliding window can be (0, k); when i==392, the area enclosed by the sliding window can be (392-k, 392); when i=100, the area enclosed by the sliding window can be (100-k/2, 100+k/2). In general, it is necessary to ensure that the size of the window along the extension direction of the target gene is the distance between k adjacent amino acids.

In one embodiment, the control group may be the control group data of the target gene (eg, TP53) as a whole. In other words, the control group data of each site are the same, and are all data where no mutation occurs at the gene site on the target gene (eg, TP53).

According to the embodiments of the present disclosure, by collecting sub-mutation data corresponding to each gene site in a sliding window manner, it is possible to obtain sub-mutation data efficiently, conveniently, concisely and clearly, thereby improving the efficiency and accuracy of gene mutation prognostic risk assessment.

According to the embodiments of the present disclosure, the control group data of each sub-mutation data can be the same, and the control data are all obtained when there is no mutation in the gene site on the target gene chain. By using the prognostic risk regression function of the proportional risk regression model between each sub-mutation data and the control data to perform a prognostic risk regression quantitative analysis, a prognostic risk value corresponding to each sub-mutation data can be obtained. The prognostic risk value corresponding to the sub-mutation data can be used as a prognostic risk value corresponding to the gene site where the gene mutation occurs, or as a prognostic risk value corresponding to the amino acid site where the gene site has mutated.

Specifically, after obtaining the sub-mutation data, regression analysis can be performed to obtain a prognostic risk value. In the disclosed embodiment, the prognostic risk value of each gene locus is characterized by the prognostic risk value of the amino acid site associated with the gene locus, and the sub-mutation data is associated with the amino acid site. The process of obtaining the prognostic risk value of the gene locus by using a proportional risk regression model to perform a quantitative prognostic risk analysis between each sub-mutation data and a control data set may include the following operations: inputting the sub-mutation data into a prognostic risk function to obtain a prognostic risk function value corresponding to each amino acid site; and using the ratio between the prognostic risk function value of each amino acid site and the benchmark prognostic risk function value as the prognostic risk value of each amino acid site, wherein the benchmark prognostic risk function value is obtained by inputting the control data set into the prognostic risk function.

According to an embodiment of the present disclosure, based on the sub-mutation data of each amino acid site and the control data set, a proportional risk regression model is used to analyze the effect of the amino acid site mutation on the prognostic risk. The proportional risk regression model is modeled based on the prognostic risk function, which can use censored data to analyze the prognostic risk rate of each factor for survival, but does not consider the distribution of survival time.

According to an embodiment of the present disclosure, the prognostic risk function may be as shown in formula (1).
h(t,X)＝h ₀ (t)exp(β ₁ X ₁ +β ₂ X ₂ +…+β _P X _P ) (1)

Wherein, h(t, X) is the prognostic risk function, the prognostic risk rate or the instantaneous mortality rate at time t, and h ₀ (t) is the baseline prognostic risk function, i.e., the prognostic risk function obtained when all variables are 0 at time t. _{X 1} , X ₂ , ..., X _p can be covariates, influencing factors or prognostic factors, and β ₁ , β ₂ ..., β _p are regression coefficients. _{X 1} , X ₂ , ..., X _p can be associated with the fifth column in FIG. 6 , and when a mutation occurs, X can be assigned a value of 1, and when no mutation occurs, X can be assigned a value of 0. The data corresponding to _{X 1} , X ₂ , ..., X _p can also be adaptively adjusted according to actual needs.

According to an embodiment of the present disclosure, the process of calculating the prognostic risk value may be as shown in formula (2).

Among them, HR represents the mutation prognostic risk value of each amino acid site. When β>0 and HR>1, the prognostic risk rate increases, indicating that variable X is a risk factor. When β<0 and HR<1, the prognostic risk rate decreases, indicating that variable X is a protective factor; when β=0 and HR=1, the prognostic risk rate remains unchanged, indicating that variable X is a risk-independent factor. _{h 0} (t) can be the baseline prognostic risk function value obtained based on the control data set, that is, the prognostic risk function value at time t when all variables are taken as 0. According to the mutation prognostic risk value characterizing the amino acid site, it can roughly correspond to the mutation prognostic risk value of the gene site.

According to the embodiments of the present disclosure, by combining the sliding window to collect sub-mutation data and using the proportional risk regression model to output the prognostic risk value of the amino acid site, it is possible to quantitatively characterize the prognostic risk of gene point mutations and improve the accuracy of gene mutation prognostic risk assessment.

FIG. 8 is a scatter plot of mutation prognostic risk values according to an embodiment of the present disclosure.

After the mutation prognosis risk value of each amino acid site is obtained by performing proportional risk regression model analysis on the sub-mutation data and the control data set of each amino acid site, a mutation prognosis risk value scatter plot can be drawn based on the prognosis risk value of each amino acid site; and the mutation prognosis risk value scatter plot can be displayed using a visualization component, and the displayed mutation prognosis risk value scatter plot can be shown in Figure 8. Among them, the visualization component can include components such as Echarts (data visualization chart library) and Highcharts (chart library).

Continuing to refer to Figure 8, the upper part of Figure 8 is a scatter plot of the mutation prognostic risk value of the amino acid site of the target gene (e.g., TP53), and the lower part is a bar graph of the number of samples in the sliding window corresponding to the amino acid site of the target gene (e.g., TP53). The X-axis represents the 393 amino acid sites of the target gene (e.g., TP53), the left Y-axis represents the logarithmic prognostic risk, and the blue dots represent the mutated amino acid sites with high credibility of the prognostic risk value (P value <0.05); other colors The dots represent amino acid sites of unknown clinical significance with low credibility (P value>0.05). The gray lines represent the 95% upper and lower confidence intervals. The larger the interval, the lower the credibility. The right Y-axis represents the number of samples, and the red and blue bars on the right represent the P value. P is between 0 and 1. The larger the P is, the darker the red is. The smaller the P is, the darker the blue is.

According to an embodiment of the present disclosure, as shown in FIG8 , the prognostic risk of point mutations obtained by the proportional risk regression model analysis can be divided into red dots and blue dots, where blue dots are points with a sufficient number of samples and high credibility, and red dots are points with an insufficient number of samples and insufficient credibility. Therefore, a deep learning method is also required to discriminate the prognostic risk of the red dots to determine whether they are protective factors or harmful factors. That is, for amino acid sites of unknown clinical significance in FIG8 , a qualitative analysis can be further performed to determine whether the prognostic risk of amino acid sites of unknown clinical significance is a high prognostic risk, a low prognostic risk, or no prognostic risk, etc. When discriminating the prognostic risk, a machine learning method can be used to discriminate the prognostic risk type of point mutations of unknown clinical significance based on features such as the effect of amino acid point mutations on protein function.

According to the embodiments of the present disclosure, by utilizing a sliding window and a prognostic risk regression function, the data of the target gene mutation data set and the control data set are analyzed and a mutation prognostic risk value scatter plot to be displayed is obtained. Based on the mutation prognostic risk value scatter plot, the prognostic risk value, 95% confidence interval and P value of each amino acid site can be obtained, which not only realizes the quantitative analysis of the prognostic risk of gene mutations, but also characterizes the prognostic risk value of amino acid site mutations in a finer granularity and more accurately. The visual display also facilitates relevant personnel to make clinical decisions based on the mutation prognostic risk value scatter plot.

According to the embodiments of the present disclosure, since the prognostic risk value of some gene loci cannot accurately reflect the clinical prognostic risk, for example, the amount of sample data obtained for some gene loci is small, the credibility is low, etc., so that the prognostic risk of clinical significance cannot be accurately obtained according to the prognostic risk value of the gene loci, it is necessary to further analyze the gene loci. Specifically, to further analyze the gene loci, it is necessary to first select the gene loci from multiple gene loci according to the prognostic risk value, and the process of selecting the gene loci from multiple gene loci according to the prognostic risk value can include the following operations: according to the prognostic risk value of each amino acid site, select the target amino acid site from the amino acid sites associated with the gene site, wherein the target amino acid site includes the target gene site.

According to the embodiments of the present disclosure, the mutation of a gene locus will cause the amino acid corresponding to the gene locus to change, and the number of amino acids is less than the number of gene loci. Therefore, using the target amino acid locus instead of the target gene locus for qualitative analysis of the prognosis risk type can simplify the workload of the analysis, improve the efficiency of the analysis of the mutation prognosis risk type, and be able to more clearly and concisely describe the prognosis risk type results of the mutation. In the process of analyzing the prognosis risk type of the mutation, the embodiments of the present disclosure perform the analysis in units of amino acid sites.

According to an embodiment of the present disclosure, the prognostic risk assessment method for gene site mutations may also include the following operations: calculating the credibility associated with the prognostic risk value of each amino acid site; The confidence associated with the risk value is compared with the confidence threshold to obtain a prognostic risk value less than the confidence threshold; the amino acid site associated with the prognostic risk value less than the confidence threshold is determined from the amino acid sites associated with the gene site to obtain the target amino acid site.

According to the embodiments of the present disclosure, since the mutation of a gene site may affect the corresponding amino acid, the confidence associated with the prognostic risk value of each gene site may be calculated by calculating the confidence associated with the prognostic risk value of each amino acid site.

As shown in Figure 8, the credibility associated with the prognostic risk value of the amino acid site can be obtained by using a tool for drawing a scatter plot. The target amino acid site can be an amino acid site of unknown clinical significance represented by a red point in Figure 8.

According to an embodiment of the present disclosure, the prognostic risk type may include high prognostic risk and low prognostic risk, etc.

According to the embodiments of the present disclosure, credibility can characterize the reliability of the prognostic risk value. By using the credibility threshold to screen the amino acid sites and taking the amino acid sites with a credibility threshold less than the credibility threshold as the target amino acid sites, it is possible to screen out the amino acid sites with low reliability, so that the qualitative analysis of the prognostic risk of the integrated classification model does not need to analyze the prognostic risk of each amino acid site, thereby improving the efficiency of the integrated classification model in performing the qualitative analysis of the prognostic risk.

According to an embodiment of the present disclosure, operation S150 may include the following operations: for the prognostic risk caused by the mutation of the target amino acid site, the prognostic risk type may be qualitatively analyzed using an integrated classification model to obtain a prognostic risk type result. Specifically, since the prognostic risk value of some gene loci cannot accurately reflect the clinical prognostic risk, for example, the sample data of some gene loci is small, the credibility is low, etc., so that the prognostic risk of clinical significance cannot be accurately obtained according to the prognostic risk value of the gene locus, it is necessary to use an integrated classification model to further qualitatively analyze the prognostic risk of this gene locus or the amino acid site associated with this gene locus.

Specifically, the process of using an integrated classification model to perform qualitative analysis of prognostic risk types may include the following operations: based on the target amino acid site, extracting gene mutation data associated with the mutation at the target amino acid site from the target gene mutation data set to obtain mutation data of the prognostic risk type to be analyzed; based on the mutation data of the prognostic risk type to be analyzed, using the integrated classification model to predict the degree of match between the prognostic risk caused by the mutation at the target amino acid site and each of the N prognostic risk types, to obtain N predicted matching values corresponding to the N prognostic risk types, where N is a positive integer, and the prognostic risk degrees between the N prognostic risk types are different from each other; and determining the prognostic risk type result based on the N predicted matching values.

According to the embodiments of the present disclosure, by using an integrated classification model to perform a qualitative assessment of amino acid sites of unknown clinical prognostic risk significance, a prognostic risk type result is obtained, which can provide reference information for clinical decision-making.

According to an embodiment of the present disclosure, based on the mutation data of the prognostic risk type to be analyzed, an integrated classification model is used to predict the degree of match between the prognostic risk caused by the mutation at the target amino acid site and each of the N prognostic risk types, and obtaining N predicted matching values corresponding to the N prognostic risk types may include the following operations: for each of the N prognostic risk types: using an integrated classification model to analyze the prognostic risk caused by the mutation at the target amino acid site and the degree of match between the prognostic risk type, to obtain multiple predicted values, wherein the integrated classification model includes multiple classifiers, each classifier can derive a predicted value; using a gradient boosting classifier to process the multiple predicted values to obtain a predicted matching value.

FIG. 9 is an architecture diagram of obtaining a predicted matching value using an integrated classification model according to an embodiment of the present disclosure.

As shown in FIG9 , the integrated classification model may include multiple classifiers, and the classifier may include at least one of the following: a random forest classifier, a Gini index random tree classifier, an entropy random tree classifier, and a gradient boosting classifier. The first extreme random tree classifier may be a Gini index random tree classifier or an entropy random tree classifier, and the second extreme random tree classifier may also be a Gini index random tree classifier or an entropy random tree classifier, but the first extreme random tree classifier and the second extreme random tree classifier preferably use different types of random tree classifiers. The Gini index random tree classifier refers to a classifier that uses the gini index as a judgment whether a node continues to split; the entropy random tree classifier refers to a classifier that uses entropy as a judgment whether a node continues to split.

The loss function of the random forest classifier is the Gini index, as shown in formula (3).

_pk can represent the probability that the prognostic risk caused by the mutation of the target amino acid site belongs to the Nth prognostic risk; the larger the Gini index, the greater the uncertainty; the smaller the Gini index, the smaller the uncertainty and the cleaner the data segmentation.

The entropy of the extreme random tree can be expressed as formula (4).

Among them, _pk can represent the probability that the prognostic risk caused by the mutation of the target amino acid site belongs to the Nth prognostic risk.

Gradient boosting is a machine learning technique for regression, classification, and ranking tasks, which can be used in the embodiments of the present disclosure to derive the degree of match between the prognostic risk caused by the mutation at the target amino acid site and each prognostic risk type.

9, when predicting the matching value, the mutation data 901 of the prognostic risk type to be analyzed can be respectively input into the random forest classifier 902, the first extreme random tree classifier 903, the second extreme random tree classifier 904, and the gradient boosting classifier 905, and the multiple predicted values obtained can be respectively output by the random forest classifier 902, the second extreme random tree classifier 904, and the gradient boosting classifier 905. The first prediction value 906, the second prediction value 907 output by the first extreme random tree classifier 903, the third prediction value 908 output by the second extreme random tree classifier 904, and the fourth prediction value 909 output by the gradient boosting classifier 905. The first prediction value 906, the second prediction value 907, the third prediction value 908 and the fourth prediction value 909 are processed by another gradient boosting classifier 910 to obtain the final required prediction matching value 911.

According to the embodiments of the present disclosure, multiple prediction values can be obtained by combining multiple classifiers to obtain an integrated classification model; and the final prediction matching value is obtained by combining multiple prediction values, thereby improving the accuracy of prognostic risk type prediction.

According to an embodiment of the present disclosure, determining a prognostic risk type result based on N predicted matching values may include the following operations: selecting the highest predicted matching value from the N predicted matching values by performing numerical comparisons between the N predicted matching values; and taking the prognostic risk type corresponding to the highest predicted matching value as the prognostic risk type result.

Exemplarily, the prognostic risk type is divided into high prognostic risk and low prognostic risk, and N is equal to 2 as an example. By calculating the matching degree between the prognostic risk caused by the mutation of the target amino acid site and the high prognostic risk type and the low prognostic risk type, the predicted matching value of the high prognostic risk type and the predicted matching value of the low prognostic risk type are obtained; by comparing the predicted matching value of the high prognostic risk type and the predicted matching value of the low prognostic risk type, when the predicted matching value of the high prognostic risk type is greater than or equal to the predicted matching value of the low prognostic risk type, the prognostic risk type result is that the prognostic risk caused by the mutation of the target amino acid site belongs to the high prognostic risk; when the predicted matching value of the high prognostic risk type is less than the predicted matching value of the low prognostic risk type, the prognostic risk type result is that the prognostic risk caused by the mutation of the target amino acid site belongs to the low prognostic risk. The above examples are only used for description and do not limit the present invention. The prognostic risk type can be adaptively divided and adjusted according to actual needs.

According to the embodiments of the present disclosure, by using an integrated classification model to perform a qualitative analysis of the prognostic risk of low-credibility prognostic risk values, it is possible to accurately discriminate the prognostic risk of new point mutations with low mutation frequency. By comparing multiple predicted matching values and taking the prognostic risk type corresponding to the highest predicted matching value as the prognostic risk type result, the final determined prognostic risk type result can be made more representative, providing a more accurate reference for clinical decision-making.

According to an embodiment of the present disclosure, for the prognostic risk type result obtained in the above process, the following operations can be performed: obtaining the prognostic risk type result; and pushing the prognostic risk type result to a target object, so that the target object generates a clinical decision according to the prognostic risk type result. The target object can be a person who makes a clinical decision, or a decision system for generating a clinical decision.

According to the embodiments of the present disclosure, by using the prognostic risk type results obtained by analyzing the integrated classification model as the basis for clinical decision-making, the efficiency of generating clinical decisions can be improved and the accuracy of clinical decisions can be increased.

In the process of training the integrated classification model in the embodiment of the present disclosure, the following operations may be adopted: obtaining a sample data set, wherein the sample data set includes first prognostic risk level data and second prognostic risk level data, and the first prognostic risk level is higher than the second prognostic risk level; randomly splitting the sample data set into an initial training sample set and a validation sample set, wherein the ratio of the number of data in the initial training sample set to the number of data in the test sample data set is a preset ratio; oversampling the initial training sample set so that the number of the first prognostic risk level data in the initial training sample set is the same as the number of the second prognostic risk level data, thereby obtaining a target training sample set; cross-validating the initial integrated classification model using the target training sample set to obtain an intermediate integrated classification model; and using the intermediate integrated classification model that meets a preset training end condition as the integrated classification model, wherein the preset training end condition includes the number of training times reaching a preset training times threshold.

According to the embodiments of the present disclosure, by oversampling the initial training sample set, the problem of unbalanced data distribution caused by random splitting of the sample data set can be avoided. By using the target training sample set to perform cross-validation training on the initial integrated classification model, the problem of low model training accuracy caused by unbalanced division of the target training sample set into training sample subsets and validation sample subsets can be eliminated, thereby improving the accuracy of the final integrated classification model.

According to an embodiment of the present disclosure, a sample data set may include multiple sample data, and the types of sample data may include at least one of the following: data on the impact of the mutated target amino acid site on protein function, data on the clinical impact of the mutated target amino acid site, and position data of the mutated target amino acid site.

Figure 10A is a schematic diagram of data obtained by using some tools to predict amino acid site mutations according to an embodiment of the present disclosure; Figure 10B is a schematic diagram of data obtained by using some tools to predict amino acid site mutations according to another embodiment of the present disclosure.

As shown in FIG. 10A to FIG. 10B , the data on the effect of the target amino acid site where the mutation occurs on the protein function can be obtained by using a test tool to analyze the effect of the amino acid point mutation on the protein function. The test tools for predicting whether the amino acid point mutation is a harmful mutation or a neutral mutation can include SIFT_score, PolyPhen2_HDIV_score, PolyPhen2_HVAR_score, LRT_score, Mutation Taster_score, Mutation Assessor_score, FATHMM_score, PROVEAN_score, VEST3_score, CADD_raw, CADD_phred, DANN_score, MetaLR_score, integrated_fitCons_score, integrated_confidence_value, GERP++_RS, fathmm-MKL_coding_score, MetaSVM_score, phyloP7way_vertebrate, phyloP20way_mammalian, phastCons7way_vertebrate, phastCons20way_mammalian, SiPhy_29way_logOdds, etc. By using the above 23 test tools for online prediction tests, 23 influence values can be obtained. In Figures 10A and 10B, only the prediction results of some tools are shown as an illustration, namely, the test results of SIFT_score, PolyPhen2_HDIV_score, PolyPhen2_HVAR_score, LRT_score, Mutation Taster_score, Mutation Assessor_score, FATHMM_score, PROVEAN_score, VEST3_score, CADD_raw, CADD_phred, DANN_score, fathmm-MKL_coding_score, MetaSVM_score, MetaLR_score, integrated_fitCons_score, integrated_confidence_value, and GERP++_RS tools are shown.

Continuing to refer to Figures 10A to 10B, the results measured by each test tool can be analyzed according to the result analysis method in the relevant technology. Exemplarily, the test results of the SIFT_score test tool are between 0 and 1, and can be divided into two ranges when analyzing the test results, 0 to 0.05 and 0.05 to 1. When the test result is between 0 and 0.05, it is recognized that the mutation at this amino acid site is harmful and will cause changes in protein function. The smaller the value, the greater the possibility of causing changes in protein function; when the test result is between 0.05 and 1, it can be considered that the mutation at this amino acid site is benign, and the closer the value is to 1, the smaller the impact on protein function. When analyzing the test results obtained by the PolyPhen2_HDIV_score, PolyPhen2_HVAR_score, LRT_score, Mutation Taster_score, Mutation Assessor_score, FATHMM_score, VEST3_score, CADD_raw, CADD_phred, DANN_score, MetaSVM_score, MetaLR_score, integrated_fitCons_score, integrated_confidence_value, and GERP++_RS test tools, the larger the value, the greater the degree of harmful effect of the mutation at the amino acid site on the protein function. The test results of PROVEAN_score can be between -14 and 14. When analyzing the test results obtained by the PROVEAN_score test tool, they can be divided into two ranges -14 to 2.5 and 2.5 to 14. When the test result is between -14 and 2.5, the mutation at this amino acid site can be considered harmful; when the test result is between 2.5 and 14, the mutation at this amino acid site can be considered neutral. When analyzing the test results obtained by the fathmm-MKL_coding_score test tool, when the test result is greater than or equal to 0.5, the mutation at this amino acid site can be considered harmful; when the test result is less than 0.5, the mutation at this amino acid site can be considered neutral.

FIG11A is a schematic diagram of data on clinical effects of amino acid site mutations according to an embodiment of the present disclosure; FIG11B is a schematic diagram of data on clinical effects of amino acid site mutations according to another embodiment of the present disclosure.

According to the embodiments of the present disclosure, the online tool VIC is used to evaluate the impact data of amino acid site variation of target genes (such as TP53) closely related to clinical conditions, such as certain mutations are signs of diagnosis of certain cancers, and certain mutations are closely related to clinical conditions. There are 12 kinds of impact data, such as mutation is a sign of good cancer prognosis, certain mutations exist in a certain pathway in cancer, etc. Each impact data can be represented by a 4-dimensional one-hot vector, and then 48-dimensional clinical impact data are obtained. Figures 11A to 11B only show the 4-dimensional one-hot vectors of some impact data.

Continuing to refer to Figures 11A to 11B, EVS_1 can represent amino acid site mutations that are recognized by food and drug regulatory agencies and are related to treatment, EVS_2 can represent amino acid site mutations that are recognized by experts and are related to diagnosis, EVS_3 can represent amino acid site mutations that are recognized by experts and are related to prognosis, and EVS_4 can represent mutation types (missense mutations, nonsense mutations, insertion and deletion mutations, and variable splicing). EVS_5, not shown, can represent the somatic mutation database COSMIC (one of the largest databases of cancer-related somatic mutation sites) and ICGC. EVS_6 can represent the score of the prediction software. EVS_7 can represent germline mutations. EVS_8 can represent cancer-related pathways. EVS_9 can represent clinically relevant mutations in public literature. EVS_10 can represent prevalent population data on the occurrence of a certain mutation in a certain type of cancer. EVS_11 can be used to represent the grade of cancer. EVS_12 can represent the type of cancer. {-1}, {0}, {1}, {2} can represent four levels, arranged from low to high, indicating the strength of evidence and clinical association, 2 is the strongest evidence, -1 is possible relevant evidence. The values of EVS_1 to 12 are one-hot types, 1 represents existence, and 0 represents non-existence. For example, EVS_1_{-1} = 1, EVS_1_{0} = 0, EVS_1_{1} = 0, EVS_1_{2} = 0, indicating that a certain mutation is a related marker for a certain cancer treatment approved by the Food and Drug Administration, but is not the strongest associated marker for a certain cancer treatment.

According to an embodiment of the present disclosure, the position data of the target amino acid site where the mutation occurs can be represented by a real number.

According to the embodiments of the present disclosure, the data of the impact of the target amino acid site that has mutated on the protein function, the data of the impact of the target amino acid site that has mutated on the clinic, and the data of the position of the target amino acid site that has mutated are spliced to obtain input data for use as initial features. It is understandable that the data of the impact of the target amino acid site that has mutated on the protein function is 23-dimensional, the data of the impact of the target amino acid site that has mutated on the clinic is 48-dimensional, and the data of the position of the target amino acid site that has mutated is 1-dimensional, so the feature data obtained after splicing can be 72-dimensional.

According to the embodiments of the present disclosure, the 72-dimensional features can be analyzed by variance analysis, and features with variance less than 1 are filtered out, and features with variance greater than 1 are retained. Since features with variance less than 1 indicate that the fluctuation between data is small, in order to better train the model, feature data with large fluctuation between data are retained.

According to an embodiment of the present disclosure, the first prognostic risk level data may refer to high prognostic risk class data, which is identified by class 1 in the embodiment of the present disclosure; the second prognostic risk level data may refer to low prognostic risk class data, which is identified by class 0 in the embodiment of the present disclosure.

According to an embodiment of the present disclosure, the process of randomly splitting a sample data set into an initial training sample set and a validation sample set, and the process of oversampling the initial training sample set can be shown in the following example. In one embodiment of the present disclosure, the number of sample data sets is 213, and the sample data sets are divided according to the ratio of 9:1 between the initial training sample set and the validation sample set, so as to obtain 191 initial training sample sets and 22 validation sample sets, and the 22 validation sample sets include 2 data of class 0. Among them, after statistics, it is found that the data distribution in the sample data set is moderately unbalanced, with 197 data in class 1 and 16 data in class 0. Therefore, before model classification, the SMOTE (Synthetic Minority Oversampling Technique) method is used to oversample the initial training sample set. Before oversampling, the ratio of class 1 to class 0 data in the initial training sample set is 177:14, and after oversampling, the ratio of class 1 to class 0 data is 177:177. Oversampling can increase the amount of class 0 data by duplicating class 0 data, ultimately making the ratio of class 1 to class 0 data 177:177.

According to the embodiments of the present disclosure, by using the impact data of the mutated target amino acid site on protein function, the clinical impact data of the mutated target amino acid site, and the position data of the mutated target amino acid site, it is possible to perform prognostic risk type analysis on amino acid sites of unknown clinical prognostic risk from multiple dimensions, thereby improving the accuracy of the integrated classification model in analyzing the prognostic risk type.

According to an embodiment of the present disclosure, cross-validation training of an initial integrated classification model using a target training sample set may include the following operations: randomly dividing the target training sample set into Q training sample subsets, where Q is a positive integer; repeatedly performing the following operations until a preset training end condition is met: selecting Q-1 training sample subsets from the Q training sample subsets to train the initial integrated classification model to obtain an intermediate integrated classification model for the next round of training, wherein the selected Q-1 training sample subsets are different from each training session to each other; using the remaining 1 training sample subset from the Q training sample subsets to test the intermediate integrated classification model for the next round of training, wherein the used training sample subsets are different from each test to each other.

According to the embodiments of the present disclosure, after the integrated classification model is trained by cross-validation, the trained integrated classification model can be verified using a validation sample set, and a ROC curve (receiver operating characteristic curve) is drawn for the output result of the integrated classification model, and the area under the ROC curve (AUC) and the accuracy (ACC) are used to evaluate the performance of the integrated classification model. AUC is the area under the ROC curve and the coordinate axis. The X-axis is the false positive rate (FPR), expressed as FPR=FP/(FP+TN), where FP (False Positive) is the number of gene mutations that are false positives, and TN (True negative) is the number of gene mutations that are true negatives. The Y-axis is the true positive rate (TPR), expressed as TPR=TP/(TP+FN), where TP (True Positive) is the number of gene mutations that are true positives, and FN (False Positive) is the number of gene mutations that are true negatives. negative) is the number of false negative gene mutations.

In a binary classification model, for the obtained model output result, when a threshold is determined, for example, the threshold is 0.5, instances greater than this threshold can be classified as positive, and instances less than this threshold can be classified as negative. Each point on the ROC curve corresponds to a threshold (the probability of each sample in the test sample being predicted as a positive class, and arranged from large to small). For a classifier, there will be a TPR and FPR under each threshold. Therefore, the determination process of the model ACC can be shown as formula (5).

FIG. 12A is a ROC curve of the integrated classification model according to an embodiment of the present disclosure.

As shown in FIG12A , the output result of the integrated classification model of the embodiment of the present disclosure may be: when the threshold is 0.5, the number of TPs is 19, the number of TNs is 1, the number of FPs is 1, and the number of FNs is 1. According to the number of FPs, FNs, TPs, and TNs, the FPR is 1/2 and the TPR is 19/20. Substituting the above-mentioned number of FPs, FNs, TPs, TNs, FPRs, and TPRs into formula (5), the ACC (test ACC) of the validation sample set is 0.909. The AUC (test AUC) of the validation sample set in FIG12A is 0.975, and the ACC (tarin ACC) of the target training sample set is 1.

FIG. 12B is a ROC curve of the related art model.

As shown in FIG12B , in order to verify that the integrated classification model of the present application can achieve a better output result, the SVM (support vector machines) in the related art was also tested. During the test, the same initial training sample set as that used to train the integrated classification model was used, and the SMOTE method was also used to oversample the initial training sample set to obtain the target training sample set, and the target training sample set was used to participate in the training. After the training was completed, the trained SVM model was verified using the verification sample set used to verify the integrated classification model. Finally, the ROC curve obtained can be shown in FIG12B . The ACC of the target training sample set is: 92.67%, the ACC of the verification sample set is: 90.91%, and the AUC of the verification sample set is: 0.88.

Continuing to refer to FIG. 12A and FIG. 12B , the ordinate value corresponding to the ROC line in FIG. 12A is higher than the ordinate value corresponding to the ROC solid line in FIG. 12B , so the area AUC under the ROC line in FIG. 12A is greater than the AUC in FIG. 12B . The larger the AUC value is, the closer it is to 1, which indicates that the accuracy of the model is higher, that is, the accuracy of the model effect of the integrated classification model in FIG. 12A is better than that of the SVM model in FIG. 12B .

The dashed lines in Figures 12A and 12B are y=x, which indicates that the false positive rate is equal to the true positive rate. The ROC line is generally above the y=x line, and the area under the line AUC can range from 0.5 to 1. The closer the AUC is to 1.0, the higher the authenticity of the model output results; when AUC is equal to 0.5, the authenticity is the lowest, indicating that The model prediction is approximately equal to random guessing and has no practical application value. So the dotted line can be a threshold. If the ROC line is lower than the blue dotted line, it means that the model has no application value.

According to the embodiments of the present disclosure, by comparing the ROC curve of the integrated classification model of the embodiments of the present disclosure with the ROC curve of the related technology, and comparing the ACC and AUC of the integrated classification model and the related technology model, it can be obtained that the AUC and ACC of the integrated classification model of the present disclosure are better than the ACC and AUC of the related technology, indicating that the integrated classification model of the embodiments of the present disclosure can more accurately classify the prognostic risk of point mutations.

According to the embodiments of the present disclosure, the initial integrated classification model is cross-validated by using the target training sample set. Since the target training sample set is divided multiple times, the training sample subset and the validation sample subset used in each training are different. This can eliminate the adverse effects caused by the unbalanced division of the target training sample set in a single division, thereby improving the accuracy of the final integrated classification model.

FIG. 13 is a method for assessing the prognostic risk of gene site mutations according to another embodiment of the present disclosure.

As shown in FIG. 13 , a prognostic risk assessment method for gene site mutation according to another embodiment may include operations S1310 to S1330 .

In operation S1310, the ICGC database and the MSK database are used to obtain TP53 point mutation data and clinical data, and preprocessing such as gene re-annotation is performed.

In operation S1320, based on the preprocessed data, a sliding window and proportional risk regression model analysis method is used to analyze the prognostic risk value of a single site mutation.

In operation S1330, based on the protein function impact score, clinical evidence score and location information, an integrated classification model method is used to predict the prognostic risk of the unknown clinical significance site.

The technical solution of the disclosed embodiment may include two parts. In the first part, a sliding window and proportional risk regression model analysis method is used to achieve quantitative characterization of the prognostic risk of gene point mutations. In the second part, a plurality of testing tools and an integrated classification model are used to achieve qualitative discrimination of the prognostic risk of gene point mutations with unknown prognostic risk. By adopting both quantitative characterization and qualitative prognostic risk analysis, the prognostic risk analysis of gene mutation sites can be achieved, providing reference information for doctors to formulate clinical plans.

According to the embodiments of the present disclosure, by using a sliding window and proportional hazard regression model analysis method, the prognostic risk of TP53 gene site mutations is analyzed using gene mutation group data composed of mutation data and clinical data and control group data to achieve quantitative analysis; based on the data on the impact of gene site mutations or amino acid site mutations on protein function, the data on the impact of gene site mutations or amino acid site mutations on clinical data, and the location data of gene site mutations or amino acid mutations, an integrated classification model is used to predict the prognostic risk of unknown clinically significant mutation sites in TP53 gene site mutations. After the risk is determined, qualitative analysis is achieved, thereby realizing the quantitative and qualitative problems of the prognostic risk of TP53 gene site mutation, refining the assessment granularity of the prognostic risk of gene mutation, improving the accuracy of assessing the prognostic risk of gene mutation, and providing a basis for clinicians to formulate clinical plans.

According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method as described above.

According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method as described above.

According to an embodiment of the present disclosure, a computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements the method as described above.

Figure 14 is a block diagram of an electronic device suitable for implementing a prognostic risk assessment method for gene site mutation according to an embodiment of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the present disclosure described herein and/or required.

As shown in FIG. 14 , the electronic device 1400 includes a computing unit 1401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1402 or a computer program loaded from a storage unit 1408 into a random access memory (RAM) 1403. In the RAM 1403, various programs and data required for the operation of the electronic device 1400 can also be stored. The computing unit 1401, the ROM 1402, and the RAM 1403 are connected to each other via a bus 1404. An input/output (I/O) interface 1405 is also connected to the bus 1404.

Multiple components in the electronic device 1400 are connected to the I/O interface 1405, including: an input unit 1406, such as a keyboard, a mouse, etc.; an output unit 1407, such as various types of displays, speakers, etc.; a storage unit 1408, such as a disk, an optical disk, etc.; and a communication unit 1409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1409 allows the electronic device 1400 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.

The computing unit 1401 may be a variety of general and/or special processing components with processing and computing capabilities. Some examples of element 1401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1401 performs the various methods and processes described above, for example, image processing methods. For example, in some embodiments, the image processing method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1408. In some embodiments, part or all of the computer program may be loaded and/or installed on the electronic device 1400 via ROM 1402 and/or communication unit 1409. When the computer program is loaded into RAM 1403 and executed by the computing unit 1401, one or more steps of the image processing method described above may be performed. Alternatively, in other embodiments, the computing unit 1401 may be configured to perform the image processing method in any other appropriate manner (e.g., by means of firmware).

Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and/or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data mining device, so that the program code, when executed by the processor or controller, implements the functions/operations specified in the flow chart and/or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include an electrical signal medium based on one or more wires. connection, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims (21)

A method for assessing the prognostic risk of gene locus mutations, comprising: Acquire an initial data set and a control data set, wherein the initial data set is obtained based on a mutation of at least one gene locus of the target gene, and the control data set is obtained based on no mutation of the gene locus of the target gene; Preprocessing the initial data set to obtain a target gene mutation data set; Determining the prognostic risk value of each gene locus in the target gene mutation data set using a proportional hazard regression model; Selecting a target gene locus according to the prognostic risk value; The prognostic risk type result of the target gene site is obtained using an integrated classification model. The method according to claim 1, wherein the initial data set includes a plurality of gene mutation data; The preprocessing of the initial data set to obtain the target gene mutation data set includes: Based on the preset gene annotation requirements, the plurality of initial gene mutation data are sorted to obtain a plurality of gene mutation data to be processed; Using a preset gene annotation statement, annotating the plurality of gene mutation data to be processed respectively to obtain a plurality of target gene mutation data, wherein each of the target gene mutation data includes a plurality of transcript information sets and a plurality of amino acid information corresponding to the gene site where the mutation occurs; and The target gene mutation data set is obtained according to the multiple target gene mutation data. The method according to claim 2, wherein obtaining the target gene mutation dataset according to the plurality of target gene mutation data comprises: Acquire target transcript information and target amino acid information associated with the target gene from a transcript database; According to the target transcript information and the target amino acid information, filtering the multiple transcript information and the multiple amino acid information included in each of the target gene mutation data, so that only the target transcript information and the target amino acid information are retained among the multiple transcript information and the multiple amino acid information of each of the target gene mutation data; and The target gene mutation data set is generated according to the plurality of target gene mutation data retaining the target transcript information and the target amino acid information. The method according to claim 1, wherein the determining the prognostic risk value of each gene locus in the target gene mutation data set using a proportional hazard regression model comprises: Based on the target gene mutation data set, determining sub-mutation data corresponding to each of the gene sites; A proportional risk regression model is used to perform a quantitative analysis of the prognostic risk between each of the sub-mutation data and the control data set to obtain the prognostic risk value of the gene locus. The method according to claim 4, wherein the target gene is linked to the target gene mutation dataset, the target gene includes m amino acid sites, each of the amino acid sites corresponds to at least one mutated gene site, and m is a positive integer; The determining, based on the target gene mutation data set, sub-mutation data corresponding to each gene site comprises: A sliding window is used to slide on the target gene, wherein the center point of the sliding window is any one of the m amino acid sites, and the size of the sliding window along the extension direction of the target gene is associated with the distance between k adjacent amino acid sites along the extension direction of the target gene, where k is a positive integer; When the sliding window slides to any one of the amino acid sites, based on the size of the sliding window along the extension direction of the target gene, k amino acid sites are selected around the amino acid site as the center to obtain the target amino acid site; and The gene mutation data linked to the target amino acid site in the target gene mutation data set is used as the sub-mutation data. The method according to claim 4, wherein the prognostic risk value of each gene locus is characterized by the prognostic risk value of an amino acid site associated with the gene locus, and the sub-mutation data is associated with the amino acid site; The method of using the proportional risk regression model to perform a quantitative prognostic risk analysis between each of the sub-mutation data and the control data set to obtain the prognostic risk value of the gene locus includes: Inputting the sub-mutation data into a prognostic risk function to obtain a prognostic risk function value corresponding to each of the amino acid sites; and The ratio between the prognostic risk function value of each amino acid site and the benchmark prognostic risk function value is used as the prognostic risk value of each amino acid site, wherein the benchmark prognostic risk function value is obtained by inputting the control data set into the prognostic risk function. The method according to claim 6, further comprising: Based on the prognostic risk value of each amino acid site, a scatter plot of mutation prognostic risk values is drawn; A visualization component is used to display the mutation prognostic risk value scatter plot. The method according to claim 7, wherein selecting a target gene locus according to the prognostic risk value comprises: According to the prognostic risk value of each of the amino acid sites, a target amino acid site is selected from the amino acid sites associated with the gene site, wherein the target amino acid site includes the target gene site. The method according to claim 8, wherein selecting a target amino acid site from the amino acid sites associated with the gene site according to the prognostic risk value of each of the amino acid sites comprises: Calculating the confidence associated with the prognostic risk value of each of the amino acid positions; Comparing the confidence associated with the prognostic risk value of each of the amino acid sites with a confidence threshold to obtain a prognostic risk value less than the confidence threshold; An amino acid site associated with a prognostic risk value less than the confidence threshold is determined from the amino acid sites associated with the gene site to obtain the target amino acid site. The method according to claim 9, wherein the step of obtaining the prognostic risk type result of the target gene locus using the integrated classification model comprises: For the prognostic risk caused by the mutation of the target amino acid site, the integrated classification model is used to perform a qualitative analysis of the prognostic risk type to obtain the prognostic risk type result. The method according to claim 10, wherein the target amino acid site is mutated The prognostic risk caused by the disease is qualitatively analyzed by using the integrated classification model, and the prognostic risk type results include: Based on the target amino acid site, extracting gene mutation data associated with the mutation at the target amino acid site from the target gene mutation data set to obtain mutation data of the prognostic risk type to be analyzed; Based on the mutation data of the prognostic risk type to be analyzed, using the integrated classification model to predict the matching degree between the prognostic risk caused by the mutation of the target amino acid site and each of the N prognostic risk types, to obtain N predicted matching values corresponding to the N prognostic risk types, where N is a positive integer, and the prognostic risk degrees between the N prognostic risk types are different from each other; and The prognostic risk type result is determined according to the N predicted matching values. The method according to claim 11, wherein the prediction of the matching degree between the prognostic risk caused by the mutation of the target amino acid site and each of the N prognostic risk types using the integrated classification model based on the mutation data of the prognostic risk type to be analyzed, and obtaining N predicted matching values corresponding to the N prognostic risk types includes: For each of the N prognostic risk types: An integrated classification model is used to analyze the matching degree between the prognostic risk caused by the mutation of the target amino acid site and the prognostic risk type to obtain multiple prediction values, wherein the integrated classification model includes multiple classifiers, each classifier can obtain one of the prediction values; The plurality of predicted values are processed using a gradient boosting classifier to obtain the predicted matching value. The method according to claim 11, wherein determining the prognostic risk type result according to the N predicted matching values comprises: Selecting a highest predicted match value from the N predicted match values by performing numerical comparison among the N predicted match values; The prognostic risk type corresponding to the highest prediction matching value is used as the prognostic risk type result. The method according to claim 13, further comprising: Obtaining the prognostic risk type result; and Push the prognostic risk type result to the target object so that the target object can Type results generate clinical decisions. The method according to claim 12, wherein the integrated classification model is trained in the following manner: Acquire a sample data set, wherein the sample data set includes first prognostic risk level data and second prognostic risk level data, and the first prognostic risk level is higher than the second prognostic risk level; The sample data set is randomly divided into an initial training sample set and a validation sample set, wherein the ratio of the number of data in the initial training sample set to the number of data in the test sample data set is a preset ratio; Performing oversampling processing on the initial training sample set so that the number of the first prognostic risk degree data and the number of the second prognostic risk degree data in the initial training sample set are the same, to obtain a target training sample set; Using the target training sample set to perform cross-validation training on the initial integrated classification model to obtain an intermediate integrated classification model; The intermediate integrated classification model that meets the preset training end condition is used as the integrated classification model, wherein the preset training end condition includes that the number of training times reaches a preset training number threshold. The method according to claim 15, wherein the cross-validation training of the initial integrated classification model using the target training sample set comprises: The target training sample set is randomly divided into Q training sample subsets, where Q is a positive integer; Repeat the following steps until the preset training end condition is met: Selecting Q-1 of the training sample subsets from the Q training sample subsets to train the initial integrated classification model to obtain an intermediate integrated classification model for the next round of training, wherein the selected Q-1 training sample subsets are different from each training to each training; The remaining 1 training sample subset among the Q training sample subsets is used to test the intermediate integrated classification model for the next round of training, wherein the training sample subset used is different from one test to another. The method according to claim 16, wherein the sample data set includes multiple sample data, and the types of the sample data include at least one of the following: data on the effect of the mutated target amino acid site on protein function, data on the clinical effect of the mutated target amino acid site, data on the effect of the mutated target amino acid site on Positional data for amino acid sites. The method according to claim 15, wherein the classifier comprises at least one of the following: a random forest classifier, a Gini index random tree classifier, an entropy random tree classifier, and a gradient boosting classifier. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are enabled to execute the method according to any one of claims 1 to 18. A computer-readable storage medium stores executable instructions, which, when executed by a processor, cause the processor to execute the method according to any one of claims 1 to 18. A computer program product comprises a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 18 is implemented.