Disease module-fused disease combination traditional chinese medicine prescription screening method and system

Abstract

The application belongs to the technical field of traditional Chinese medicine, and relates to a disease module fused disease condition combined traditional Chinese medicine prescription screening method and system. The method comprises the following steps: integrating medical knowledge graphs related to traditional Chinese and Western medicine to form data resources for a target disease; obtaining seed genes of the target disease according to the data resources for the target disease; mapping the seed genes of the target disease to a protein interaction network, expanding the genes of the target disease, and forming a disease module; determining a target point set of the target disease according to the disease module and a target point library; and performing drug matching and prescription matching according to the target point set of the target disease to obtain a traditional Chinese medicine prescription screening result. The application fuses phenotype-genotype multi-level knowledge to perform traditional Chinese medicine prescription screening, and the formed traditional Chinese medicine prescription has good correlation with the target disease at a mechanism level and has good interpretability.

Description

Technical Field

[0001] This invention belongs to the field of traditional Chinese medicine technology, specifically relating to a method and system for screening prescriptions for traditional Chinese medicine that integrates disease modules. Background Technology

[0002] With the development of artificial intelligence technology, combining artificial intelligence technology to screen new Chinese medicine drugs and construct new prescriptions is an inevitable path for the intelligent development of modern Chinese medicine.

[0003] In recent years, Tang Guoxin et al. (CN116469527A) invented an optimization recommendation method for traditional Chinese medicine prescriptions. This method includes two steps: identifying target concurrent symptoms from a pre-determined set of concurrent symptoms and determining the target drug set for the target concurrent symptoms. By obtaining the basic coefficients and weights of each drug set corresponding to the target concurrent symptoms, the selection probability of each drug set is determined based on the basic coefficients and weights of each drug set. Then, the target drug set recommended to be added to the current traditional Chinese medicine prescription is selected from each drug set. This scheme provides a technical solution to the problem of long time consumption and low efficiency in the determination of optimization strategies for traditional Chinese medicine prescription composition involved in existing traditional Chinese medicine diagnosis and treatment. Yang Kuo et al. (CN114783571A) designed a method and system for optimizing dynamic TCM diagnosis and treatment plans based on deep reinforcement learning. This method trains a diagnosis and treatment optimization model by combining a reinforcement learning model of Q-Learning and convolutional neural networks, and a reinforcement learning model that integrates fully connected networks and LSTM neural networks, to obtain the best patient diagnosis and treatment plan. By designing an offline virtual environment, reward function, state transition prediction, and diagnosis and treatment plan clustering analysis, a dynamic TCM diagnosis and treatment plan optimization model based on deep reinforcement learning is constructed.

[0004] Although technologies for screening and optimizing traditional Chinese medicine prescriptions have been proposed in recent years, these technologies still have the following shortcomings: On the one hand, existing technologies lack interpretability. Most existing technologies are built based on deep learning models, and the overall structure of the models is mainly black boxes, so they cannot provide explicit explanations for the predicted drugs. On the other hand, existing technologies only stay at the phenotypic level, focusing more on the correlation analysis between symptoms and traditional Chinese medicine, and lacking an understanding of the mechanism at the mechanistic level and the use of mechanistic knowledge in the process of formulating prescriptions. Summary of the Invention

[0005] To enhance the R&D capabilities of new traditional Chinese medicine (TCM) drugs, the development of TCM drug screening technologies based on artificial intelligence is essential. However, current technologies suffer from drawbacks such as the unexplainable prediction process and the lack of support from knowledge related to the mechanisms of TCM. Therefore, this invention addresses the current lack of explainable drug screening technologies from a mechanistic perspective by proposing a method for constructing disease modules at the molecular level and screening TCM prescriptions, thereby improving the interpretability and mechanistic rationality of existing technologies.

[0006] The technical solution adopted in this invention is as follows:

[0007] A method for screening symptoms combined with traditional Chinese medicine prescriptions by integrating a disease module includes the following steps:

[0008] Integrate medical knowledge graphs related to traditional Chinese and Western medicine to form data resources targeting specific diseases;

[0009] Based on data resources for the target disease, obtain the seed genes of the target disease;

[0010] The seed genes of the target disease are mapped onto a protein-protein interaction network, and the genes of the target disease are expanded to form a disease module.

[0011] Based on the disease module and target library, determine the target set for the target disease;

[0012] Drug matching and prescription matching are performed based on the target set of the target disease to obtain the results of traditional Chinese medicine prescription screening.

[0013] Furthermore, the data resources for the target disease include symptom-gene relationships, traditional Chinese medicine-prescription relationships, disease-gene relationships, drug-target relationships, protein-protein relationships, prescription-efficacy relationships, and symptom-syndrome relationships.

[0014] Furthermore, obtaining the seed gene of the target disease based on data resources for the target disease includes:

[0015] Based on user needs, using information about the target disease as input, and based on the integrated data resources for the target disease, the nodes related to the target disease and their neighboring information are associated.

[0016] The data of different types of relationships are merged, the gene-related relationships are extracted, and all genes related to the target disease are merged. The merged gene list is used as the seed gene for the target disease.

[0017] Furthermore, the process of mapping the seed genes of the target disease onto a protein-protein interaction network, and expanding the genes of the target disease to form a disease module, includes:

[0018] Seed genes are mapped onto a protein-protein interaction network, and first-order and second-order neighbor nodes are incorporated to form a disease module related to the target symptom; the disease module is represented as follows:

[0019] M=S∪N(S)∪N 2(S)

[0020] Where S represents the set of seed genes; N(S) represents the set of all first-order neighbors of the seed gene; N 2(S) Let represent the set of second-order neighbors of all first-order neighbors;

[0021] Evaluate the topological centralization and functional homogeneity of the disease module M;

[0022] The topological centrality is measured using the connection density D within the disease module M:

[0023]

[0024] Among them, |E M | is the number of edges in the disease module M; |V M | is the number of nodes in the disease module M;

[0025] Based on the functional significance p-values ​​of genes in disease module M, the score HS of the functional homogeneity is defined as follows:

[0026]

[0027] Where T represents the set of functions involved by genes in disease module M; P t It is the significance p-value of function t; |T| represents the total number of functions in which genes in disease module M are involved.

[0028] Furthermore, determining the target set of the target disease based on the disease module and the target library includes: taking the seed gene, first-order neighbor, and second-order neighbor of the disease module as input, combining them with the target library to determine the target, and identifying direct and indirect targets.

[0029] Furthermore, the drug matching is based on three methods to associate targets with drugs, namely:

[0030] 1) By directly linking traditional Chinese medicines with targets based on the target disease, a drug ranking list is formed by ranking the frequency of the associated drugs.

[0031] 2) Following the path of "target-symptom-traditional Chinese medicine", the disease targets of the target disease are first associated with the symptom-gene relationship, and the frequency of the associated symptoms is used as the sorting to form a symptom ranking list; then, the relationship between traditional Chinese medicine and symptoms is combined and the frequency of the associated traditional Chinese medicine is used as the sorting to form a traditional Chinese medicine ranking result.

[0032] 3) Following the path of "target-disease-symptom-drug", the target of the disease is first associated with the disease through the disease-gene relationship, and the frequency of the associated diseases is used as the sorting to form a disease ranking list; based on the top-ranked diseases, the symptoms are associated through the disease-symptom relationship, and the frequency of the associated symptoms is used as the sorting to form a symptom prediction list; and based on the symptom-traditional Chinese medicine relationship, the symptoms are associated, and the frequency of the associated traditional Chinese medicine is used as the sorting to form a traditional Chinese medicine prediction list.

[0033] Furthermore, the prescription matching includes: taking the associated drug list as input, matching it with the associated traditional Chinese medicine-prescription relationship, using the Jaccard similarity between sets as the evaluation method, then giving the relevant traditional Chinese medicine prescriptions an evaluation score, and using the ranking of the evaluation scores as the prescription matching result.

[0034] A disease-integrated system for screening traditional Chinese medicine prescriptions, comprising:

[0035] The data resource integration module for the target disease is used to integrate medical knowledge graphs related to traditional Chinese and Western medicine to form data resources for the target disease; and to obtain the seed genes of the target disease based on the data resources for the target disease.

[0036] The disease module construction module is used to map the seed genes of the target disease to the protein-protein interaction network, expand the genes of the target disease to form a disease module, and determine the target set of the target disease based on the disease module and the target library.

[0037] The disease-based interpretable TCM prescription screening module is used to match drugs and prescriptions based on the target set of the target disease to obtain TCM prescription screening results.

[0038] The beneficial effects of this invention are as follows:

[0039] (1) The disease module-based disease-based traditional Chinese medicine screening technology constructed in this invention has good interpretability and can interpret the path of the generated traditional Chinese medicine prescription throughout the entire process, and can provide reasoning path at each stage.

[0040] (2) This invention integrates phenotypic-genotypic multi-level knowledge for screening and forming Chinese medicine prescriptions, and can fully consider molecular-level information for screening new Chinese medicine drugs, so that the formed Chinese medicine prescription has a good correlation with the target disease at the mechanism level.

[0041] (3) This invention constructs a disease module, integrates multi-source data such as symptoms-genes and traditional Chinese medicine-prescriptions, and screens traditional Chinese medicine prescriptions through three paths (direct target, symptom association, and disease association), highlighting interpretability and multi-dimensional associations at the molecular level.

[0042] (4) The present invention adopts a modular design, expands seed genes through disease modules (incorporating first-order and second-order neighbors), and combines functional enrichment analysis and topological evaluation, making it more systematic; through multi-path association such as "target-symptom-drug", it enhances the comprehensiveness of screening, and from the perspective of the data resources on which it relies, this technology can avoid the result deviation caused by the current unsatisfactory drug target quality.

[0043] (5) This invention focuses on screening Chinese medicine prescriptions rather than predicting general drug combinations. By combining molecular networks with Chinese medicine disease theory, it focuses on the interpretability at the mechanism level. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the model structure of the disease-integrated traditional Chinese medicine prescription screening method of the present invention. Detailed Implementation

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0046] This invention focuses on constructing disease modules, forming disease modules based on molecular networks, and generating effective drug screening results, such as... Figure 1 As shown, it specifically includes the following three modules, which will be introduced in detail below.

[0047] (1) Data resource integration module for target diseases

[0048] This module serves as the starting point and input module for this technology. It integrates data resources based on the user's target symptom and existing biomedical datasets to form the input and background network used to subsequently generate disease modules and prescription screening results. First, this technology integrates existing medical knowledge graphs related to both traditional Chinese and Western medicine to form data resources for the target symptom, including symptom-gene relationships, traditional Chinese medicine-prescription relationships, disease-gene relationships, drug-target relationships, protein-protein relationships, prescription-efficacy relationships, and symptom-syndrome relationships. This integration of data resources provides a background network for subsequent model predictions. The background network refers to the knowledge structure or information framework that serves as the foundation for model prediction. It integrates various relationships and data from the medical knowledge graph, providing a comprehensive perspective for understanding and analyzing the complexity of the target symptom. The background network not only covers relationships between symptoms, genes, drugs, and diseases, but also includes the unique perspective of integrating traditional Chinese and Western medicine, helping the model more accurately identify and associate different medical entities and concepts during prediction.

[0049] Furthermore, based on the integration of relevant data resources, and according to user needs, using information about the target disease / symptom / syndrome as input, and based on the formed TCM and Western medicine knowledge graph resources, nodes related to the user's target disease and their neighboring information are associated, linking to the neighboring information of relevant nodes. The associated relationships, including disease-gene relationships, disease-symptom relationships, symptom-gene relationships, syndrome-symptom relationships, and symptom-drug relationships, are fused. Fusion refers to integrating different types of relationship data together, placing these relationships within a unified framework to better understand and analyze their interactions. Through the fused knowledge graph, information about nodes related to the target disease and their neighboring nodes is associated. This association is not limited to direct relationships but also includes indirect relationships, expanding the understanding of the target disease through information from neighboring nodes. Gene-related relationships are extracted, and then all genes related to the current disease (i.e., the target disease) are merged. The merged gene list serves as the input to the disease module construction module (i.e., the seed genes of the disease).

[0050] (2) Disease Module Construction Module

[0051] The data resource integration module for the target disease described above can obtain the seed genes and background networks of the user's target disease. The second module uses these as input to form a disease module, generate relevant evaluations, and identify a set of targets. The content of this module will lay the foundation for the subsequent drug screening module.

[0052] The first part concerns disease module identification. This invention uses the previously formed seed genes as input, mapping them to the protein-protein interaction network of the background network. Since different input seed genes result in different connections within the protein network after mapping, the key to studying the target disease at the molecular level lies in identifying and forming components more strongly associated with the user's input target disease based on the protein-protein interaction network containing the seed genes. Therefore, this technology expands upon the disease module network by incorporating first- and second-order neighbor information, thereby forming disease modules related to the user's target disease. After forming the disease module, the topological concentration and functional homogeneity of the disease module are evaluated, including topological analysis results, functional enrichment analysis results, and randomized controlled trials results, to assess the current topological concentration and functional homogeneity of the disease module. The core idea of ​​this module is to consider other genes potentially related to the seed genes in the protein-protein interaction network as genes possibly related to the current disease. This allows for the expansion of the target disease genes based on the original seed genes, thereby forming a more comprehensive disease module at the molecular level.

[0053] To form a disease module related to the target symptom, firstly, based on the protein-protein interaction network of the seed gene, first-order and second-order neighbor nodes are incorporated:

[0054] M=S∪N(S)∪N 2(S)

[0055] Where S represents the set of seed genes; N(S) represents the set of all first-order neighbors of the seed gene; N 2(S) This represents the set of second-order neighbors of all first-order neighbors. First-order neighbors are other nodes in the network that are directly connected to a node (i.e., nodes that can be accessed through a direct connection edge), while second-order neighbors are nodes that can be accessed in two steps.

[0056] By unifying these nodes (seed gene nodes, first-order neighbor nodes, and second-order neighbor nodes) into a set M, a disease module is formed.

[0057] The topological centrality of disease module M can be measured using the connection density (D) within disease module M:

[0058]

[0059] Among them, |E M | is the number of edges in the disease module M; |V M | represents the number of nodes in disease module M. Higher connection density indicates stronger interactions between genes within disease module M.

[0060] The functional homogeneity of disease module M was assessed using functional enrichment analysis. Based on the functional significance p-values ​​of genes in disease module M, a homogeneity score (HS) was defined.

[0061]

[0062] Where T represents the set of functions involved by genes in disease module M; P t The p-value represents the significance of function t; |T| represents the total number of functions involved by genes in disease module M. A higher homogeneity score indicates that genes within disease module M are more functionally consistent.

[0063] After identifying and obtaining the disease module, the next step is to determine the target, which lays the groundwork for subsequent screening of traditional Chinese medicine prescriptions. The target determination module needs to combine the relevant genes (seed genes, first-order neighbors, and second-order neighbors) retained after identification by the disease module with a target database to determine which are targets and which are direct and indirect targets. Specifically, the relevant genes of the disease module are used as input, combined with the existing target database for judgment, and relevant targets overlapping with those in the existing database are retained. After screening by the target database, if the relevant gene is the original seed gene of the user-input disease, then the relevant gene is a direct target; if the relevant gene is a new gene expanded and included through the previous step of disease module identification, then the relevant gene is an indirect target. Therefore, after this step, this technology can determine the direct and indirect targets related to the target disease, thus providing a reference for subsequent prescription screening and disease understanding.

[0064] (3) Explanable Traditional Chinese Medicine Prescription Screening Module Based on Symptom Combination

[0065] In this module, the input is the disease target information obtained in the previous module. This module extracts these targets and their relationships from the related medical knowledge graph, including relationships between traditional Chinese medicine (TCM) and targets, diseases and genes, diseases and symptoms, symptoms and genes, symptoms and drugs, and TCM and prescriptions. Based on these data resources and the disease target list generated in the previous module, this module first associates targets with drugs, and then matches the associated drugs with prescriptions to generate prescription screening results.

[0066] The first step is the drug matching process. This module will associate targets with drugs using three methods:

[0067] 1) By directly linking traditional Chinese medicines with targets based on the target disease, a drug ranking list is formed by ranking the frequency of the associated drugs.

[0068] 2) Following the "target-symptom-traditional Chinese medicine" path, the disease targets of the target disease are first associated with the symptom-gene relationship, and the frequency of the associated symptoms is used as the sorting to form a symptom ranking list. Then, based on the associated symptom list, the traditional Chinese medicine-symptom relationship is combined and associated, and the frequency of the associated traditional Chinese medicine is used as the sorting to form a traditional Chinese medicine ranking result.

[0069] 3) Following the path of "target-disease-symptom-drug," the target of the symptom is first associated with existing diseases through the disease-gene relationship, and then sorted by the frequency of the associated diseases to form a disease ranking list. Based on the association results of this step, it is possible to observe not only whether the order of the disease association list has a good correlation with the user's input symptom, but also whether the top-ranked related diseases have a strong correlation with the current user's input symptom, which will also provide a reference for drug recommendation. Based on the top-ranked diseases, symptoms are associated through the disease-symptom relationship, and sorted by the frequency of the associated symptoms to form a symptom prediction list. Then, based on the symptom-traditional Chinese medicine relationship, a traditional Chinese medicine prediction list is formed, sorted by the frequency of the associated traditional Chinese medicines.

[0070] The above "target-disease-symptom-drug" path can be formally described as follows: Assume the target set for the target disease is T, the existing disease set is Dis = {d1, d2, ..., dm}, and the gene set associated with each disease di is G(di). By calculating the intersection of the target set T and the gene set G(di) for each disease, the disease association frequency F(di) is defined as:

[0071] F(di)=|T∩G(di)|

[0072] Then, sort the diseases in descending order of F(di) to form a disease ranking list:

[0073] Dis′=Sort(Dis,F(di),descending)

[0074] Sort means sorting, and descending means sorting in descending order.

[0075] For the top-ranked disease set Dis′ k ={d′1,d′2,…,d′ k (Select the first k diseases), assuming each disease has a d′ i The associated symptom set is S(d′) i The set of symptoms associated with all selected diseases is as follows:

[0076]

[0077] For each symptom s∈S, its frequency F(s) is determined by the number of diseases associated with that symptom:

[0078] F(s)=|{d′ i ∈Dis′ k |s∈S(d′ i )}|

[0079] Then, sort them in descending order of F(s) to form a symptom prediction list:

[0080] S′=Sort(S,F(s),descending)

[0081] For the symptom set S′ ranked first l ={s1′,s2′,…,sl′} (select the first l symptoms), assuming each symptom s′ j The associated set of traditional Chinese medicines is H(s′). j The set of Chinese herbal medicines associated with all selected symptoms is as follows:

[0082]

[0083] For each Chinese medicine h∈H, its frequency F(h) is determined by the number of symptoms associated with that Chinese medicine:

[0084] F(h)=|{s′ j ∈S′ l |h∈H(s′ j )}|

[0085] Then, sort them in descending order according to F(h) to form a list of predicted Chinese medicines:

[0086] H′=Sort(H,F(h),descending)

[0087] Based on the three design paths mentioned above, a drug ranking list related to the three paths can be generated for user reference.

[0088] The next step is the prescription matching process. The list of drugs associated in the previous step is used as input (using the top 20 drugs in the ranking), and then matched with the associated Chinese medicine-prescription relationships. The matching method uses Jaccard similarity between sets as the evaluation. Then, the relevant Chinese medicine prescriptions can be given an evaluation score, and the ranking of the evaluation score similarity is used as the prescription matching result.

[0089] Through the above-described method, this technology is implemented through three modules: a seed gene list related to the disease, a disease module based on the seed gene list and related statistical information, and direct and indirect disease targets obtained after screening. After passing through a drug screening module, a drug matching list and a prescription matching list for three pathways related to the target disease are obtained. This invention enables the mining of molecular information about a given disease, combined with existing medical resource databases, and then allows for drug mining and prescription matching based on this molecular information. Furthermore, the entire process of this technology is transparent and produces interpretable results.

[0090] The key point of this invention is:

[0091] (1) Data resource integration technology for target diseases used as input for traditional Chinese medicine screening models;

[0092] (2) Disease module construction technology combined with disease module detection technology;

[0093] (3) Screening technology for interpretable Chinese medicine prescriptions based on symptoms.

[0094] The following example, using the combination of diabetic nephropathy and spleen-kidney yang deficiency syndrome, illustrates the process of forming drug screening results through the above method, and provides sample results.

[0095] Diabetic kidney disease (DKD) is a chronic kidney disease caused by diabetes mellitus and is one of the most common microvascular complications in people with diabetes. The main clinical manifestations of DKD are proteinuria, edema, elevated blood pressure, decreased urine output, and progressive renal function impairment. In traditional Chinese medicine, DKD falls under the category of a variant of "Xiao Ke Bing" (wasting and thirsting disease), and many physicians classify it as "Xiao Ke Bing Nephropathy." Spleen and kidney yang deficiency is a common syndrome in the middle and late stages of DKD.

[0096] The first step involved retrieving relevant information about the disease and forming a knowledge graph. Based on medical literature and guidelines, symptoms of spleen and kidney yang deficiency included "aversion to cold and cold limbs," "lower back and knee weakness," "deep, slow, and weak pulse," "limb edema," and "turbid urine." By searching relevant databases, genes (i.e., seed genes) related to spleen and kidney yang deficiency and diabetic nephropathy were identified, including ADA, DCLRE1C, IL7R, MVK, and RAG2. Subsequently, a disease module was constructed using disease module detection technology. The core genes involved in the constructed disease module included JAK2, IL10, IL18, STAT1, EPOR, IL12RB1, CSF3, IL2RG, and STAT4. Then, through indirect association via the "target-disease-symptom-drug" pathway, drugs were ultimately derived. Symptoms involved in this association included: severe muscle pain, liver yang dizziness, migraine, forgetfulness, chest pain radiating to the back, cold joint pain, spontaneous sweating, memory loss, and anemia. The resulting drug order was: Chuanxiong (Ligusticum striatum), Gegen (Pueraria lobata), Wuzhuyu (Evodia rutaecarpa), Jili (Tribulus terrestris), Niuhuang (Calculus bovis), Fenge (Pueraria lobata), Baishao (Paeonia lactiflora), Jue Mingzi (Cassia tora), Xiangru (Elsholtzia ciliata), and Guanghuoxiang (Pogostemon cablin). The resulting prescription matching results were: modified Shengyang Yiwei Tang, Yishen Huashi Keli (Kidney-Nourishing and Dampness-Eliminating Granules), Jiajian Shentang Fang (Modified Kidney-Sugar Formula), Jiangtang Yishen Wan (Kidney-Nourishing and Blood-Nourishing Pill), and Jiawei Danggui Liuhuang Tang (Modified Danggui Liuhuang Decoction). The top 10 prescriptions for diabetes and kidney deficiency showed a precision rate of 50% and a recall rate of 5 / 96 = 5.2%. This demonstrates that through a series of calculations and associations, this method can ultimately yield prescriptions related to diabetes and spleen-kidney yang deficiency, showcasing the effectiveness of this method.

[0097] Another embodiment of the present invention provides a computer device (computer, server, smartphone, etc.) including a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program including instructions for performing the steps of the method of the present invention.

[0098] Another embodiment of the present invention provides a computer-readable storage medium (such as ROM / RAM, disk, optical disk) storing a computer program that, when executed by a computer, implements the various steps of the method of the present invention.

[0099] The specific embodiments of the present invention disclosed above are intended to help understand the content of the present invention and to implement it accordingly. Those skilled in the art will understand that various substitutions, changes, and modifications are possible without departing from the spirit and scope of the present invention. The present invention should not be limited to the content disclosed in the embodiments of this specification; the scope of protection of the present invention is defined by the claims.

Claims

1. A method for screening traditional Chinese medicine prescriptions based on symptoms combined with a disease module, characterized in that, Includes the following steps: Integrate medical knowledge graphs related to traditional Chinese and Western medicine to form data resources targeting specific diseases; Based on data resources for the target disease, obtain the seed genes of the target disease; The seed genes of the target disease are mapped onto a protein-protein interaction network, and the genes of the target disease are expanded to form a disease module. Based on the disease module and target library, determine the target set for the target disease; Drug matching and prescription matching are performed based on the target set of the target disease to obtain the results of traditional Chinese medicine prescription screening.

2. The method according to claim 1, characterized in that, The data resources targeting specific diseases include symptom-gene relationships, traditional Chinese medicine-formula relationships, disease-gene relationships, drug-target relationships, protein-protein relationships, prescription-efficacy relationships, and symptom-syndrome relationships.

3. The method according to claim 1, characterized in that, The step of obtaining the seed genes for the target disease based on data resources for the target disease includes: Based on user needs, using information about the target disease as input, and based on the integrated data resources for the target disease, the nodes related to the target disease and their neighboring information are associated. The data of different types of relationships are merged, the gene-related relationships are extracted, and all genes related to the target disease are merged. The merged gene list is used as the seed gene for the target disease.

4. The method according to claim 1, characterized in that, The process of mapping seed genes of the target disease onto a protein-protein interaction network, and expanding the genes of the target disease to form a disease module, includes: Seed genes are mapped onto a protein-protein interaction network, and first-order and second-order neighbor nodes are incorporated to form a disease module related to the target symptom; the disease module is represented as follows: M=S∪N(S)∪N 2(S) Where S represents the set of seed genes; N(S) represents the set of all first-order neighbors of the seed gene; N 2(S) Let represent the set of second-order neighbors of all first-order neighbors; Evaluate the topological centralization and functional homogeneity of the disease module M; The topological centrality is measured using the connection density D within the disease module M: Among them, |E M | is the number of edges in the disease module M; |V M | is the number of nodes in the disease module M; Based on the functional significance p-values ​​of genes in disease module M, the score HS of the functional homogeneity is defined as follows: Where T represents the set of functions involved by genes in disease module M; P t It is the significance p-value of function t; |T| represents the total number of functions in which genes in disease module M are involved.

5. The method according to claim 4, characterized in that, The step of determining the target set of the target disease based on the disease module and the target library includes: taking the seed gene, first-order neighbor, and second-order neighbor of the disease module as input, combining them with the target library to determine the target, and identifying direct and indirect targets.

6. The method according to claim 1, characterized in that, The drug matching is based on three methods to associate targets with drugs, namely: 1) By directly linking the "Chinese medicine-target" relationship, related Chinese medicines are associated with the target set of the target disease, and the frequency of the associated drugs is used as the sorting to form a drug sorting list; 2) Following the path of "target-symptom-traditional Chinese medicine", the disease targets of the target disease are first associated with the symptom-gene relationship, and the frequency of the associated symptoms is used as the sorting to form a symptom ranking list; then, the relationship between traditional Chinese medicine and symptoms is combined and the frequency of the associated traditional Chinese medicine is used as the sorting to form a traditional Chinese medicine ranking result. 3) Following the path of "target-disease-symptom-drug", the target of the disease is first associated with the disease through the disease-gene relationship, and the frequency of the associated diseases is used as the sorting to form a disease ranking list; based on the top-ranked diseases, the symptoms are associated through the disease-symptom relationship, and the frequency of the associated symptoms is used as the sorting to form a symptom prediction list; and based on the symptom-traditional Chinese medicine relationship, the symptoms are associated, and the frequency of the associated traditional Chinese medicine is used as the sorting to form a traditional Chinese medicine prediction list.

7. The method according to claim 1, characterized in that, The prescription matching includes: taking the associated drug list as input, matching it with the associated traditional Chinese medicine-prescription relationship, using the Jaccard similarity between sets as the evaluation method, then giving the relevant traditional Chinese medicine prescriptions an evaluation score, and using the ranking of the evaluation scores as the prescription matching result.

8. A disease-integrated system for screening traditional Chinese medicine prescriptions, characterized in that, include: The data resource integration module for target diseases is used to integrate medical knowledge graphs related to traditional Chinese and Western medicine to form data resources for the target diseases. Based on data resources for the target disease, obtain the seed genes of the target disease; The disease module construction module is used to map the seed genes of the target disease to the protein-protein interaction network, expand the genes of the target disease to form a disease module, and determine the target set of the target disease based on the disease module and the target library. The disease-based interpretable TCM prescription screening module is used to match drugs and prescriptions based on the target set of the target disease to obtain TCM prescription screening results.

9. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program configured to be executed by the processor, the computer program including instructions for performing the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a computer, implements the method according to any one of claims 1 to 7.

Images