CN107103207B - Accurate medical knowledge search system based on case multigroup variation characteristics and implementation method - Google Patents

Abstract

The invention discloses a precise medical knowledge searching system based on multi-case-science variation characteristics and an implementation method thereof, wherein the system comprises: the method is realized by the steps of establishing a multi-precise medical knowledge base based on a multiomic variation-intervention response association model, extracting multiple groups of mathematical variation characteristics of a new case, establishing a matching algorithm between the new case and the model to generate an analysis report of a case matching system, updating data of the knowledge base and self-evolving of the matching algorithm. The invention systematically integrates the correlation between known omic variation and intervention response, and integrates the intervention response and omic variation information of different levels and sources into a knowledge base by defining a generalized framework of a plurality of groups of mathematical variation-intervention response correlation model types.

Description

Precision medicine knowledge search system and implementation based on multi-omics variation characteristics of cases method

technical field

The invention belongs to the field of medicine and health industry, and relates to an implementation method of a precise medical knowledge search system, in particular to an implementation method of a precise medical knowledge search system based on multi-omics variation characteristics of cases.

Background technique

Precision medicine relies on the classification of biomarkers for disease risk, prognosis and treatment response. The rapid development of omics technology has greatly enriched the number of biomarkers at the molecular level, providing a more comprehensive and detailed basis for diagnosing disease, judging disease staging, or evaluating the safety and efficacy of new therapies in target populations.

At present, the association information of "molecular-level markers or pathomic variation characteristics-intervention response (including drug response)" can mainly be obtained from several channels such as companion diagnostics, high-throughput drug screening experiments at the cell line level, and precision medicine clinical trials. . The association information provided by the companion diagnosis is obtained under the observation of a large sample population, that is, the population level, and the information is directly and easily obtained. However, the correlation information provided by cell line drug screening experiments and precision medicine experiments needs to be processed on the original information, and the correlation between molecular-level variation and intervention response can be established by extracting multi-omics variation features from omics data. Therefore, the current situation that different sources and types of association information are mixed and indistinguishable has increased the difficulty for many clinicians to interpret the physiological significance of omics variation characteristics and utilize the clinical value.

In addition, the integration and clinical translation of omics data also need to consider the issue of data stability, experimental platform (such as different laboratories or institutions), observation scale (such as cell line level, tissue level, individual level, etc.), observation method ( Factors such as transcriptome level, proteome level, or genome level, etc.), observation methods (such as single nucleotide polymorphism chip, next-generation sequencing technology, etc.) and other factors may cause the observed instability of the same biomarker behavior. Therefore, how to integrate these related information to the greatest extent and make them play the best role still needs to be solved urgently.

SUMMARY OF THE INVENTION

The purpose of the present invention is to use the observable individual multi-omics variation characteristics to quickly search the knowledge base for the multi-omics variation-intervention response correlation models that are successfully matched with new cases, and to compare the intervention strategies corresponding to all the successfully matched models and whether they are successful or not. The record of the successful response is presented to the user in an easy-to-read and tightly integrated form, and the present invention is achieved through the following technical solutions:

The invention discloses a precise medical knowledge search system based on the multi-omics variation characteristics of cases. The system includes:

A precision medicine knowledge base for collecting multi-omics variation-intervention response correlation models, which realizes the collection and integration of "omics variation characteristics-intervention response" information at different levels;

An optimizable matching algorithm for judging whether the case matches the model in the knowledge base and the degree of matching;

The evaluation algorithm of the matching algorithm is used to compare the clustering results of the knowledge base model by evaluating the matching algorithm and the results obtained by the model according to the label classification of the intervention response, so as to evaluate the advantages and disadvantages of the matching algorithm and continuously optimize the algorithm;

The report directly generated by the search system includes case omics analysis data and system search results, which is used to provide doctors with physiological significance of omics data and assist in the formulation of treatment plans.

As a further improvement, the different levels described in the present invention include population level, individual level, tissue level and cell line level.

The invention also discloses a method for realizing a precise medical knowledge search system based on the multi-omics variation characteristics of cases, which is realized by the following steps:

1) Establish a multi-precision medicine knowledge base based on a multi-omics variant-intervention response correlation model;

2) When a new case appears, extract the multi-omics variation characteristics of the new case;

3) Establish a matching algorithm between new cases and models (known multi-omics variation characteristics-intervention response associations);

4), generate the analysis report of the case matching system;

5) Data update of knowledge base and self-evolution of matching algorithm.

As a further improvement, in step 1) of the present invention, the multi-omics variation information includes single-base mutations (single nucleotide polymorphisms and base indels) in the genomic regions with active transcription, chromosomal mutations (such as gene fusion) and the benchmark gene expression level used to judge whether the gene expression is abnormal.

As a further improvement, in step 1) of the present invention, a multi-omics variation-intervention response correlation model is a set of "companion diagnostic correlation models" with companion diagnostic drug response annotations and multi-omics variation characteristics, or "Cell-line association models" that include drug response information and multi-omic variation characteristics in drug screening experiments, or clinically observed "case association models" that include intervention response outcomes and multi-omics variation characteristics, or include drug screening "Individualized Disease Model Association Models" for Outcome Information and Multi-omics Variation Signatures. The individualized models include but are not limited to PDX mice and PDO organoid models.

As a further improvement, in step 2) of the present invention, the multi-omics variation characteristics include single base mutation, chromosome structural variation, and abnormal gene expression information in the genomic region with active transcription.

As a further improvement, in step 2) of the present invention, a set of standardized omics data analysis process is established to extract multi-omics variation, and the whole process from sample collection, sequencing, data analysis, to knowledge base matching is carried out for quality control and Warranty.

As a further improvement, in step 3) of the present invention, the search system provides an initial matching algorithm and an evaluation method for the matching algorithm, and the evaluation method will be based on the clustering performance of the association models in the knowledge base using different matching algorithms To evaluate whether the existing algorithm is better than the new algorithm, decide whether the algorithm needs to be upgraded and optimized.

As a further improvement, in step 4) of the present invention, the report is divided into two parts: the first part is the statistical information display of the physiologically related multi-omics variation characteristics of cases, from single base mutation, chromosomal variation The second part is to show the matching evidence and medication of the model in descending order according to the similarity between the model and the case in the system after completing the search of the knowledge base. information.

As a further improvement, in step 5) of the present invention, after the case completes step 2) omics feature extraction, track the drug treatment effect of the case, add the case data as a case model to the precision medicine knowledge base, and expand the knowledge Improve the coverage of the database and improve the matching accuracy of the knowledge base; when no matching association model is found in the knowledge base, the treatment can be directly based on the doctor's experience, and at the same time, the case can be developed to establish an individualized disease model, and the treatment effect and individualization of the case can be tracked. Based on the drug test results of the disease model, the corresponding "case association model" or "individualized disease model association model" is constructed and added to the precision medicine knowledge base.

The advantages of the present invention are:

1) The present invention has a wide search range and can retrieve correlation models under different observation scales. The present invention systematically integrates the known associations between omics variation and intervention response, and by defining a generalized framework of multi-omics variation-intervention response association model classes, the intervention response and omics variation of different levels and sources Information is consolidated into a knowledge base.

2) The available matching features and matching strategies of the present invention are abundant. On the one hand, the cooperative matching of multi-omics variation characteristics from multiple aspects such as single base variation, chromosomal variation, and differentially expressed genes ensures the reliability of the matching results and reduces the noise in the correlation analysis between single variation types and physiological phenotypes. On the other hand, the present invention provides specific and optimizable matching strategies for intervention response models of different scales in the knowledge base, and provides multi-angle evidence support for the relationship between cases and intervention responses through the association model.

3) The present invention has self-evolution ability. This capability is manifested in two aspects: First, the number of models in the precision medicine knowledge base will continue to expand with the operation of the search system. After a new case is entered, the system will record the multi-omics variation characteristics of the case, and combine the follow-up treatment plan and intervention response results of the case or the medication results of the individualized disease model of the case to generate an association model of the case and add it to the multi-precision medicine knowledge base. Second, the matching algorithm of the system can be continuously optimized. The present invention establishes a corresponding evaluation method for the matching algorithm. Once the matching algorithm is updated, the new matching algorithm can be used to re-cluster the models in the knowledge base, compared with the classification based on the intervention response labels, and whether the system needs to be updated by evaluating whether the new algorithm is better than the existing algorithm.

4) The present invention fills the gap between the link of omics variation information extraction and the link of clinical guidance medication, and assists clinical staff in systematic interpretation of the physiological significance of omics variation and mining of clinical value.

Description of drawings

FIG. 1 is a schematic diagram of the implementation flow of the technical solution of the present invention.

Detailed ways

The present invention establishes a precise medical knowledge search system based on the multi-omics variation collaborative matching method of individual cases. The system of the present invention: 1. It includes a precision medicine knowledge base. The knowledge base realizes the collection and integration of "omics variation characteristics-intervention response" information at different levels (population level, individual level, tissue level, cell line level, etc.) by collecting multi-omics variation-intervention response correlation models. Individual cases entered into the system can be used as new models for the expansion of the knowledge base; 2. It includes an optimizable matching algorithm. The initial matching algorithm provided by the system does not maximize the advantage of rich omics variation, but the present invention provides an evaluation method for the matching algorithm, by evaluating the clustering result of the matching algorithm on the knowledge base model, and the model according to the model. By comparing the results obtained from the label classification of the intervention response, the pros and cons of the matching algorithm can be evaluated, and the algorithm can be continuously optimized; 3. The search system directly generates an easy-to-read report containing the case study data and system search results, which can Provide physicians with the physiological significance of omics data and the formulation of adjuvant treatment plans.

The basic model of this invention is as follows: 1. Establish a multi-precision medicine knowledge base based on a multi-omics variation-intervention response correlation model. The multi-omics variation information includes three aspects: single base mutation (single nucleotide polymorphism and base insertion and deletion), chromosomal variation (such as gene fusion) and the benchmark gene expression level used to judge whether the gene expression is abnormal. A multi-omics variant-intervention response association model can be a set of "companion diagnostic association models" with companion diagnostic drug response annotations and multi-omic variation characteristics; it can also be a drug screening experiment that includes drug response information and multi-omics variation. It can be a “cell line association model” of characteristics; it can also be a clinically observed “case association model” that includes intervention response results and multi-omics variation characteristics; it can also be an “individual” that includes drug screening results information and multi-omics variation characteristics. "Disease-associated models" (including but not limited to PDX mice, PDO organoid models). 2. When a new case appears, extract the multi-omics variation characteristics of the new case (including but not limited to single base mutation, chromosomal structural variation, and gene expression profile information). Establish a standardized omics data analysis process to extract multi-omics variants, and conduct quality control and quality assurance throughout the entire process from sample collection, sequencing, data analysis, to knowledge base matching. 3. Establish a matching algorithm between new cases and association models. The search system provides a starting matching algorithm and an evaluation method for the matching algorithm. The evaluation method will evaluate whether the existing algorithm is better than the new algorithm according to the clustering performance of the association models in the knowledge base using different matching algorithms, and decide whether to Algorithm upgrade and optimization. Fourth, generate personalized reports of cases. The report is divided into two parts: the first part is the statistical information of the multi-omics variation characteristics related to the physiology of the cases, and the omics variation information of the diseased tissue is given from the aspects of single base mutation, chromosomal variation and differentially expressed genes; Part of it is to display the matching evidence and medication information of the model in descending order according to the similarity between the model and the case in the system after completing the search of the knowledge base. 5. If the case does not match the existing model, the drug can be used directly according to the doctor's experience. At the same time, an individualized disease treatment model based on the case can be developed for drug screening. Disease Association Models", added to Knowledge Base.

Fig. 1 is a schematic diagram of the implementation flow of the technical solution of the present invention, and the specific implementation steps are as follows:

1) Build a precision medicine knowledge base based on a multi-omics variation-intervention response association model: build intervention response models at different scales (including but not limited to population level, individual level, tissue level, and cell line level), including but not limited to "Population Omics Variation Characteristics-Intervention Response", "Individual Case Omics Variation Characteristics-Intervention Response", "Omics Variation Characteristics of Individualized Disease Models (such as PDX Mice and PDO Models, etc.)-Intervention Response", "Cell Lines From the perspectives of "omics variation characteristics-intervention response", collect multi-omics variation characteristics and corresponding intervention and intervention response information. The data in this knowledge base is obtained through web crawling, public database download, and local data import (cases and individualized disease models). The obtained data needs to undergo word segmentation, semantic analysis, regular matching and other technologies to extract core keywords and data, and then convert the format to the original information. The data of the same type of association model in the database has a unified information storage format;

2) Build a process for extracting multi-omics variation characteristics of cases: Build a bioinformatics analysis process based on next-generation sequencing technology, and extract single-base mutations, genome structure mutations and transcriptional level expressions that are closely related to physiological changes from omics data. Abnormal genes, as multi-omics variant signatures of cases, are used to match models in the Multi-Omics Variation Signature Database. The data analysis process of the cases uses strict quality control. When normal control samples are available, the normal samples and known disease-omics variation information are used to screen the case omics variation, and the multi-omics variation characteristics of the cases are increased. Reliability of physiological phenotype associations;

3) Realize the case-model multi-omics variation collaborative matching algorithm: The precision medicine knowledge base integrates the variation feature information of the association model from multiple data sources and multi-omics perspectives. When a case completes the extraction of multi-omics variation characteristics and enters the case matching system, it is necessary to match the case with the model according to the type of the model in the knowledge base. When matching with a specific association model, it is necessary to use different methods for different omics variation characteristics to match and score the variation characteristics extracted from the cases and the variation characteristics of the model. The scoring generates the matching total score of the case-drug response model according to the formula, and judges whether the case and the model can be matched according to the total score;

4) Generate the analysis report of the case matching system: the report is divided into two levels: the first level: the omics information report of the individual case. Including but not limited to raw data sequencing quality information, data analysis process introduction, statistical information of multi-omics variation characteristics; second layer: matching results between cases and models in the precision medicine knowledge base. According to the search results, the model's intervention strategy, response results, and matching evidence are displayed in descending order of the similarity between the model and the case in the system. The second layer provides easy-to-read information of "individual case omics variation characteristics-model omics variation characteristics-intervention response", which provides the potential intervention response information of cases to assist doctors in interpreting the physiological meaning of omics variation characteristics and mining omics the clinical value of the data;

5) The update of the search system: the update of the system is divided into two parts: the data update of the knowledge base and the self-evolution of the matching algorithm.

1. Update of the knowledge base: When a case is matched with the model in the knowledge base, track the drug treatment effect of the case, add the case data as a case model to the precision medicine knowledge base, expand the coverage of the knowledge base and improve the matching of the knowledge base precision. When no matching association model is found in the knowledge base, the treatment can be directly based on the doctor's experience. At the same time, the case can be developed to establish an individualized disease model (such as PDX mouse or PDO organoid model, etc.), and the case intervention response results and individuals can be tracked. Based on the drug test results of the disease model, the corresponding case association model or individual disease association model is constructed and added to the precision medicine knowledge base.

2. Self-evolution of matching algorithm: The system establishes an evaluation method for comparing the advantages and disadvantages of new and old matching algorithms to optimize the system matching algorithm. When the system is put into operation, an initial matching algorithm to be optimized is provided first. As new cases expand, the models in the precision medicine knowledge base will continue to grow, providing resources for optimizing matching algorithms. According to the classification of the response of the models in the knowledge base to the intervention, the present invention can randomly select M correlation models, use the old and new matching algorithms to score between the selected models, and obtain two similarity scoring matrices composed of these models. . By further clustering the matrix, the model classification obtained by the old and new matching algorithms can be obtained and compared with the actual classification results based on the drug response information, so as to judge whether the new algorithm performs better and can replace the current algorithm of the system.

The technical scheme of the present invention is further described below by specific embodiments:

Example 1: A Rapid Matching System for Cancer Cases Based on Transcriptome Variation Signatures of Cases

This embodiment consists of five steps:

1) Construction of multiple precision medicine knowledge bases: The knowledge base uses the association model as the storage object, from the list of companion diagnostic drugs approved by the US Food and Drug Administration (FDA), the precise cancer medical information provided by My Cancer Genome, and the Sanger Institute The three data sources of the GDSC database collect drug response information associated with multi-omics variability signatures. Companion diagnostic drugs and My Cancer Genome provide population-level omics variation signature-drug response information, and GDSC database provides cell line-specific omics variation signature-drug response information. Data in different formats are managed uniformly through the naming method provided by the international standard database. In this example, single-base mutations from different sources are mapped to the corresponding names in the COSMIC database, and the names in the database are used as the standard output. Likewise, gene names use NCBI's entrez ID as a standard, and disease names use OMIM ID as a standard.

2) Extraction of multi-omics variation characteristics of cases: build a bioinformatics analysis process based on transcriptome sequencing (RNA-Seq) data, and extract single-base mutations, chromosome structure mutations and mutations that are closely related to physiological changes from transcriptome data. Genes with abnormal expression at the transcript level were used as multi-omics variant signatures of cases for matching with models in the Multi-Omics Variation Signature Database.

In this example, the variant extraction process can be divided into the following parts: RNA-Seq data preprocessing, single base mutation detection (single nucleotide polymorphism, small fragment indels), chromosomal structural variation detection ( Gene fusion), detection of gene expression and abnormally expressed genes, and visual display of results.

1. RNA-Seq data preprocessing:

The raw data was checked for data quality using quality control tools, and the detected data was subsequently excised using de-linker software to excise linker sequences and low-quality bases in the reads. The cleaned reads are used for subsequent sequence alignments. Here, this example uses the Fast Short Fragment Alignment software and the human genome as the reference genome for alignment.

2. Detection of single base mutations in cases:

This example follows the RNA-seq variant calling best practice process provided by GATK at this step (http://gatkforums.broadinstitute.org/gatk/discussion/3892/the-gatk-best-practices-for-variant-calling- on-rnaseq-in-full-detail) to operate. First, the redundant reads are removed from the files obtained by the comparison in 1., and then the reads are trimmed, and the reads are disassembled by exon segments, and base correction is performed. Sex and single-nucleotide indels were detected, and finally, the detected single-base variants were annotated and filtered using the human genome variation database resources and variant annotation software.

3. Detection of chromosomal aberrations in cases:

Structural variations detected by transcriptome sequencing data are mainly gene fusions. The gene fusion events that can be seen on the transcriptome are detected by gene fusion software for the alignment results in 1. here.

4. Detection of gene expression:

This step also uses the alignment file in 1. as the input file for the fragment assembly software for transcript assembly and expression calculation. In this example we only consider the case where no paracancerous tissue is provided nor in the public cancer transcriptome database.

5. Visual display of case data results:

The overall multi-omic variation profile of individual cases is shown in circle plots. The circle diagram is composed of four parts from the inside to the outside. The innermost part shows the location of the gene fusion event, then the location of the single base mutation event, followed by the expression of the gene in the entire transcriptome, and the outermost layer is Annotated chromosomal location information.

All kinds of statistical charts generated in the analysis process, such as scatter plots, histograms, pie charts, etc., are visualized and output through the statistical software R.

3) Realization of case-model multi-omics variation cooperative matching algorithm: The multi-omics variation characteristic database integrates the variation characteristic information of the association model from multiple data sources and multi-omics perspectives. When a case completes the extraction of multi-omics variation characteristics and enters the case matching system, it is necessary to provide a case-model matching algorithm according to the type of model in the database.

In this example, the knowledge base provides three types of models: 1. Companion diagnostic association models; 2. Cell line association models; 3. Case association models.

The intervention result given by the population-level association model is usually the effect of one or several specific omics variation characteristics on the drug response in a large population sample. Therefore, the strategy adopted by the example for this model is that if the case and a population model have exactly the same omics variation characteristics during the comparison, the reported case is successfully matched with the population model, otherwise the matching fails.

Both cell line-level association models and individual-level association models provide complete information on single-base mutations, chromosomal structural mutations, and gene expression profiles. Therefore, this example adopts a similarity scoring method that combines these three aspects of information to measure the similarity between cases and models. The difference between using the association model at the cell line level and the association model at the individual level to match cases is that the threshold parameters that ultimately determine whether the matching is successful are different. The following are the implementation steps of the scoring method:

1. For single-base mutations: This example uses the DANN method to measure the functional importance of single-base mutations in cases and models, respectively, for the DANN of sites with significant single-base functional mutations on each gene in the case and model. The values are summed to measure the physiological impact of single-base functional mutations in the gene. The similarity score of the functional mutation of the gene in the case and the model can be obtained by the formula 1-|C _snv -M _snv |/Max{C _snv ,M _snv }, where C _snv is the functional mutation impact value of a gene in the case , M _snv is the functional mutation effect value of the model. This score can be used as an indicator V1 to measure the similarity of gene function between cases and models.

2. For chromosomal structural variation: There is no method to directly measure the degree of physiological impact of gene fusion. Considering that structural variation usually has a very serious impact on the physiological function of genes, this example uses a custom indicator V2 (0 or 1) to measure the similarity between cases and samples in gene fusion events. If a gene fusion occurs in both cases and models, or no gene fusion occurs, the V2 value is 1; otherwise, the V2 value is 0.

3. For abnormally expressed genes: This example defines an indicator V3 to measure abnormal gene expression, the formula is V3=1-|C _exp -M _exp |/Max{C _exp ,M _exp }, where C _exp and M _exp are the expression levels of a gene in cases and models after the expression profile is normalized.

In this example, considering that abnormal gene expression reflects variation at the transcriptional level, and single-base mutation or chromosomal structural variation reflects variation on the genome, it is necessary to integrate the effects of both when integrating these indicators. In the example, the final similarity score between the case and the model for a certain gene is defined as V=Min{V3*V1, V3*V2}, where V1, V2, V3 are the three similarity indicators mentioned above in the specification. For a specific gene, if the similarity score is higher than 0.5, the gene is considered to be consistent in the case and model. When more than half of the genes in the cases are consistent with their performance in the model, the case and the model are considered to be successfully matched, otherwise the matching is considered unsuccessful.

4) Generate an analysis report according to the matching results of the cases:

The analysis report display is mainly divided into two parts: individual case information and knowledge base search result display.

In this example, the display of individual case information includes:

1. Basic information of sequencing samples (including sample name, sample delivery time, sequencing time, sequencer model, sample label, and data saturation evaluation parameters);

2. The overall display of the omics data, and the statistical information of the transcriptome sequencing data (including the number of original reads of the sample, the number of reads after cleaning, the number of reads compared to the reference genome, and the number of reads in the specific alignment);

3. The histogram of the expression distribution of the detected genes, and the chart of the differentially expressed genes;

4. Quantitative statistics of single base variation and structural variation on the genome and interpretation of the variation file format;

5. Raw data QC report location, expression file location of genes and transcripts, file location of differentially expressed genes, file location of single base variation information, file location of gene fusion information.

The knowledge base search result display includes:

1. Basic information of the matched model (model type, original data source, model name, disease name, etc.);

2. Evidence to support the model on case matching (type of index, index name, index measurement value, etc.) on model and case matching;

3. The clinical drug reference information of the matched model (drug name, whether the model responds to the drug, etc.)

5) Self-evolution of the search system:

1. Update of the precision medicine knowledge base: Track the cases entered into the knowledge base for analysis, and establish a case-omics variation characteristic-intervention response correlation model according to the medical compliance treatment effect and long-term outcome of the cases, and join the knowledge base. For cases where no matching model is found in the knowledge base for the first time, consider establishing an individualized disease model (PDX mouse model or PDO organoid model), and establish an individualized disease model based on the response of the in vitro individualized disease model to different drugs. Variation signature-drug response association model, added to knowledge base.

2. Self-evolution of matching algorithm: When the number of models of a certain type in the knowledge base in the search system accumulates to a certain value, M models of this type can be randomly selected and classified according to their responses to drugs for targeting this type of model. The evaluation of the matching algorithm. When a matching algorithm for a new case and this type of model is implemented, the evaluation results of the new matching algorithm and the old matching algorithm can be compared. If the new method is more consistent with the classification according to the response of the drug, it means that the new matching algorithm has better application effect in the real situation, and the matching algorithm is updated; otherwise, the original algorithm is better, and the update algorithm is abandoned.

The foregoing enumerations are merely specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there can be many modifications, and all modifications that those of ordinary skill in the art can directly derive or associate from the disclosure of the present invention should be considered as the protection scope of the present invention.

Claims (9)

1. a precision medicine knowledge search system based on case multi-omics variation characteristics, is characterized in that, described system comprises: A precision medicine knowledge base for collecting multi-omics variation-intervention response correlation models, which realizes the collection and integration of "omics variation characteristics-intervention response" information at different levels; An optimizable matching algorithm for judging whether the case matches the model in the knowledge base and the degree of matching; The evaluation algorithm of the matching algorithm is used to compare the clustering results of the knowledge base model by evaluating the matching algorithm and the results obtained by the model according to the label classification of the intervention response, so as to evaluate the advantages and disadvantages of the matching algorithm and continuously optimize the algorithm; The report directly generated by the search system includes case omics analysis data and system search results, which is used to provide doctors with physiological significance of omics data and assist in the formulation of treatment plans. 2 . The precision medicine knowledge search system based on multi-omics variation characteristics of cases according to claim 1 , wherein the different levels include population level, individual level, tissue level and cell line level. 3 . 3. a realization method of the precision medicine knowledge search system based on case multi-omics variation feature as claimed in claim 1 or 2, is characterized in that, described realization method is realized by the following steps: 1) Establish a precision medicine knowledge base based on a multi-omics variant-intervention response correlation model; 2) When a new case appears, extract the multi-omics variation characteristics of the new case; 3) Establish a matching algorithm between new cases and multi-omics variation-intervention response association models; 4), generate the analysis report of the case matching system; 5) Data update of knowledge base and self-evolution of matching algorithm. 4. The realization method of the precision medicine knowledge search system based on case multi-omics variation characteristics according to claim 3, is characterized in that, in described step 1), multi-omics variation information comprises in the genomic region that transcription is active. Single base mutation, chromosomal variation and the benchmark gene expression amount used to judge whether the gene is abnormally expressed. 5. the realization method of the precision medicine knowledge search system based on case multi-omics variation feature according to claim 3, is characterized in that, in described step 1), a multi-omics variation-intervention response correlation model is a A "companion diagnostic association model" with accompanying diagnostic drug response annotations and multi-omics variation signatures, or a "cell line association model" that includes drug response information and multi-omics variation signatures in drug screening experiments, or clinically observed A "case association model" that includes intervention response results and multi-omics variation characteristics, or an "individualized disease model association model" that includes drug screening results information and multi-omics variation characteristics. 6. the realization method of the precision medicine knowledge search system based on case multi-omics variation feature according to claim 4 or 5, is characterized in that, in described step 2), establish a set of standardized omics data analysis process Extract multi-omics variants, and perform quality control and quality assurance throughout the entire process from sample collection, sequencing, data analysis, to knowledge base matching. 7. the realization method of the precision medicine knowledge search system based on case multi-omics variation feature according to claim 6, is characterized in that, in described step 3), search system provides an initial matching algorithm and is aimed at matching. The evaluation method of the algorithm. The evaluation method will evaluate whether the existing algorithm is better than the new algorithm according to the clustering performance of the association model in the knowledge base using different matching algorithms, and decide whether to upgrade and optimize the algorithm. 8. The realization method of the precision medicine knowledge search system based on case multi-omics variation feature according to claim 4 or 5 or 7, is characterized in that, in described step 4), described report is divided into two parts : The first part is the statistical information display of the physiologically related multi-omics variation characteristics of the cases, and the omics variation information of the diseased tissue is given from the aspects of single base mutation, chromosomal variation and differentially expressed genes; After the knowledge base is searched, the matching evidence and medication information of the model are displayed in descending order according to the similarity between the model and the case in the system. 9. according to the realization method of the accurate medicine knowledge search system based on case multi-omics variation feature according to claim 3, it is characterized in that, in described step 5), when case matches model in knowledge base, Track the drug treatment effect of cases, add the multi-omics variation characteristic data of actual cases and drug effect as a "case association model" into the precision medicine knowledge base, expand the coverage of the knowledge base and improve the matching accuracy of the knowledge base; when the knowledge base When no matching association model is found in the search, the treatment can be directly based on the doctor's experience, and the case can be developed to establish an individualized disease model, and the response results of the case intervention and the test drug results of the individual disease model can be tracked, and the corresponding "case association model" can be constructed. Or "individualized disease model association model" to join the precision medicine knowledge base.

Images