A method for optimizing and evaluating compound Chinese medicine prescriptions

Abstract

The present invention relates to a method for optimizing and evaluating a compound Chinese medicine prescription. The present invention independently constructs a prediction model for the safety of Chinese medicinal materials, takes the theory of "monarch, minister, assistant and envoy" of traditional Chinese medicine as a guide, and uses multidisciplinary technology to provide a method for optimizing a compound Chinese medicine prescription. This method takes into account the safety of Chinese medicine while retaining the essence of Chinese medicine to the greatest extent. At the same time, the present invention provides an evaluation method for comparing the efficacy of the optimized compound Chinese medicine prescription with the original prescription based on bioinformatics and network pharmacology. The method can quickly obtain the chemical components verified by experiments in the compound Chinese medicine, with fast acquisition speed and high accuracy. The method enables scientific researchers to avoid a large amount of search and sorting work, and the labor cost is low. In addition, the present invention can quickly judge the application value of the optimized prescription, which provides a new idea and reference for later researchers to evaluate the efficacy of the optimized compound Chinese medicine prescription.

Description

Technical Field

[0001] The present invention belongs to the field of traditional Chinese medicine development and bioinformatics technology. Specifically, the present invention relates to a method for optimizing the composition of a compound traditional Chinese medicine by using a deep learning algorithm model, a database and text mining, and evaluating the optimized composition of the compound traditional Chinese medicine by using bioinformatics and network pharmacology. Background Art

[0002] Traditional Chinese medicine is an important part of my country's traditional ethnic medicine industry, with a long history and strong ethnic characteristics. As a new way and method to actively develop traditional Chinese medicine, the secondary development of traditional Chinese medicine is crucial to the modernization of traditional Chinese medicine. The 2020 edition of the "Chinese Pharmacopoeia" contains a total of 2,711 kinds of traditional Chinese medicine. Among the finished prescriptions, compound traditional Chinese medicines with more than 10 Chinese medicinal materials account for about 35.3%. Therefore, focusing on the optimization of compound traditional Chinese medicine formulas is an effective way to drive the leapfrog development of the traditional Chinese medicine industry through innovation. The optimization of the formula of compound traditional Chinese medicines that have passed clinical verification, efficacy certification, and have a certain research and development foundation not only conforms to the basic theory of traditional Chinese medicine in formula and clinical practice, and maintains the essential characteristics of traditional Chinese medicine, but also its clinical effectiveness can also improve the pertinence of new drug development and reduce blindness. The optimization of compound traditional Chinese medicine formulas can systematically identify the chemical composition of traditional Chinese medicines, pharmacological substances, harmful substances, and the processes of the main components in the body, thereby improving the quality control level of drugs. At the same time, the application of artificial intelligence and big data to optimize the formulation of compound Chinese medicine has given traditional Chinese medicine a new development direction, increased the scientific and technological connotation of Chinese medicine, improved the quality and technical content of Chinese medicine, and laid a solid foundation for the innovation of Chinese medicine and the standardization and scientificization of Chinese medicine. In addition, the value of traditional drugs is mainly reflected through clinical experiments, which requires a lot of time and effort, which is a great challenge for Chinese medicine with many ingredients and complex mechanisms. Therefore, after optimizing the formulation of clinically verified compound Chinese medicine, there is an urgent need for a fast and effective method to compare the efficacy of the optimized formulation with the original formulation. Summary of the invention

[0003] The purpose of the present invention is to provide a method for optimizing the formulation of a compound Chinese medicine and a method for evaluating the efficacy of an optimized formulation of a compound Chinese medicine (hereinafter referred to as the improved formulation) and an original formulation (hereinafter referred to as the original formulation), so as to provide a basis and reference for the optimization and evaluation of the formulation of other compound Chinese medicines.

[0004] The present invention provides a method for optimizing and evaluating a compound Chinese medicine prescription. Through a database, based on computational analysis, an optimized compound Chinese medicine prescription is obtained and correlation and experimental verification are performed. At the same time, through text mining and computational analysis, the efficacy of the improved compound Chinese medicine prescription is evaluated. Among them, steps S1-S4 are a method for optimizing the compound Chinese medicine prescription, and steps S5-S8 are a method for evaluating the efficacy of the improved compound Chinese medicine prescription.

[0005] In a first aspect, the present invention provides a method for optimizing a compound Chinese medicine prescription, comprising the following steps:

[0006] S1: Build a Chinese herbal medicine library and collect the chemical components of all Chinese herbal medicines in the library;

[0007] S2: Based on the chemical composition of all Chinese medicinal materials, the improved formula of compound Chinese medicine is obtained through computational analysis;

[0008] S3: Evaluate the efficacy of the improved prescription by calculating the correlation between the target set of the improved prescription and the disease gene set;

[0009] S4: Verify the efficacy of the improved formula through animal experiments.

[0010] Preferably, in the Chinese medicinal materials library described in step S1, the Chinese medicinal materials are derived from the Chinese Pharmacopoeia, and the chemical components of the Chinese medicinal materials are derived from known public databases.

[0011] Furthermore, step S2 includes: based on the chemical composition of all Chinese medicinal materials in step S1, using the deep neural network (DNN) algorithm to construct a Chinese medicinal material safety prediction model; using the DNN model to predict the safety of the adjuvant and the adjuvant of the compound Chinese medicine, respectively, to obtain the safety prediction results of the adjuvant and the adjuvant; if the safety prediction of the adjuvant and the adjuvant is extremely low, low or relatively low, then directly eliminate the Chinese medicinal material to obtain an improved prescription.

[0012] Furthermore, step S2 includes: if the safety prediction of the adjuvant and the adjuvant is high, an improved prescription is obtained after comprehensive analysis of the ADME properties of the Chinese medicinal material components and target network correlation analysis.

[0013] Furthermore, the ADME property analysis includes: based on the chemical composition of the Chinese medicinal materials in step S1, standardizing the component names and eliminating duplicate data, obtaining the chemical composition of each Chinese medicinal material in a certain compound Chinese medicine formula; screening, eliminating and sorting the chemical components of adjuvants and adjuvants according to ADME properties, and the evaluation indicators include intestinal absorption, oral availability and Lipinski rule.

[0014] Preferably, if the unique chemical components of adjuvants and guiding drugs in the compound Chinese medicine formula are the largest (i.e., the chemical components of adjuvants and guiding drugs are less covered by the chemical components of monarch drugs and ministerial drugs), this Chinese medicinal material is directly retained because it may have relatively unique medicinal properties.

[0015] Furthermore, the target network correlation analysis includes: obtaining the target of each Chinese medicinal material in the compound Chinese medicine formula through text mining, standardizing the target name and eliminating duplicate data to obtain the corresponding target set; based on the target set, constructing the target network of the original compound Chinese medicine formula; retaining all targets of the monarch and ministerial drugs in the compound Chinese medicine formula, and randomly eliminating the targets of the adjuvant or guiding drugs as the improved formula target set; after randomly eliminating the targets of the adjuvant or guiding drugs, comparing the centrality of the eliminated targets in the original formula network and the importance in the key sub-network (importance of degree centrality>importance of proximity centrality=importance of MCODE score).

[0016] Furthermore, after screening and sorting by comprehensive ADME property analysis and target network correlation analysis, the Chinese medicinal materials that need to be eliminated are finally obtained. Preferably, after screening, the Chinese medicinal material with the highest score can be eliminated as an improved formula, or the top two Chinese medicinal materials can be eliminated at the same time as an improved formula.

[0017] Furthermore, step S3 includes: collecting one or more disease genes related to the compound Chinese medicine, and obtaining the disease gene set after artificial correction and removal of duplicate data; calculating the correlation between the target set of the original and improved compound Chinese medicine and the disease gene set based on the network propagation perspective, and evaluating the efficacy of the original and improved formulas on the disease respectively.

[0018] Preferably, the disease gene is derived from a known public database.

[0019] Preferably, the random walk algorithm is used to calculate the correlation between the target set of the improved prescription and the original prescription and the disease gene set, and a Z-score>3 indicates that the drug and the disease are significantly correlated.

[0020] Furthermore, step S4 includes: detecting corresponding observation indicators of the above-mentioned improved formula and original formula respectively, and verifying the efficacy of the improved formula through animal experiments.

[0021] In a second aspect, the present invention provides a method for evaluating the efficacy of an improved compound Chinese medicine prescription, comprising the following steps:

[0022] S5: Construct a chemical component library of compound Chinese medicine (i.e., non-predicted, derived from known research chemical components);

[0023] S6: Based on the compound Chinese medicine chemical component library, the active targets of the chemical components of the original compound Chinese medicine and its improved formula were obtained, and the active targets were enriched and analyzed and compared;

[0024] S7: Obtain disease genes related to the compound Chinese medicine, and based on the active targets and disease genes in step S6, evaluate the efficacy of the modified formula on the disease by calculating the network proximity of the active targets and disease genes between the original compound Chinese medicine and its modified formula;

[0025] S8: Test the corresponding observation indicators of the improved prescription and the original prescription respectively to verify the efficacy of the improved prescription.

[0026] Preferably, step S5 collects the chemical components of the compound Chinese medicine through literature or Xingdou Cloud knowledge graph text mining tools, standardizes the names of the components and removes duplicate data, and then obtains the corresponding chemical component libraries.

[0027] Preferably, the method for obtaining the chemical component target in step S6 is a method well known in the art, for example, it can be obtained through public databases PubChem (https: / / pubchem.ncbi.nlm.nih.gov / ), Drugbank (https: / / go.drugbank.com / ), ChEMBL (https: / / www.ebi.ac.uk / chembl / ), HERB (http: / / herb.ac.cn / ), Pubmed (https: / / pubmed.ncbi.nlm.nih.gov / ), CNKI (https: / / www.cnki.net / ), and the species is selected as "Homo sapiens".

[0028] Preferably, the step S7 of obtaining disease genes related to the compound Chinese medicine includes: obtaining one or more disease genes related to the compound Chinese medicine, the disease genes are derived from a known database, and the disease genes are obtained after artificial correction.

[0029] The method of the present invention covers the following technical points:

[0030] 1. Construct a DNN prediction model and use it to predict the safety of compound Chinese medicines based on the principle of “monarch, minister, assistant and envoy” compatibility;

[0031] 2. Predict the ADME properties of the chemical components of each Chinese medicinal material in the compound Chinese medicine, compare and analyze the Chinese medicinal materials that need to be eliminated and rank them;

[0032] 3. Analyze the network centrality and target importance of the targets to be eliminated for each Chinese medicinal material in the compound Chinese medicine, compare the Chinese medicinal materials that need to be eliminated and rank them;

[0033] 4. Compare and analyze the target set and disease gene set of the compound Chinese medicine before and after improvement, and use the random walk algorithm to compare the effects or efficacy of the improved compound.

[0034] 5. Carry out enrichment analysis on the active targets of the chemical components of the compound Chinese medicine and its improved formula, and compare the pathways and tissue sites of action of the compound Chinese medicine and its improved formula;

[0035] 6. Conduct network proximity analysis on the active targets of the chemical components and disease genes of the compound Chinese medicine and its improved formula, and compare the efficacy of the compound Chinese medicine and its improved formula.

[0036] The present invention independently constructs a prediction model for the safety of Chinese medicinal materials, and takes the theory of "monarch, minister, assistant and messenger" of traditional Chinese medicine as a guide, and uses multidisciplinary technologies such as bioinformatics, network pharmacology, chemoinformatics and systems biology to provide a method for optimizing the compound Chinese medicine formula. This method takes into account the safety of Chinese medicine while retaining the essence of Chinese medicine to the greatest extent. This method is of great significance to the development of Chinese medicine, can enhance the scientific and technological connotation and international market competitiveness of Chinese medicine, and promote the sustainable and healthy development of Chinese medicine. At the same time, the present invention provides an evaluation method for comparing the efficacy of the optimized compound Chinese medicine formula with the original formula based on bioinformatics and network pharmacology. This method can quickly obtain the chemical components verified by experiments in the compound Chinese medicine, with fast acquisition speed and high accuracy. This method enables scientific researchers to avoid a large amount of search and sorting work, and the labor cost is low. In addition, this method can quickly judge the application value of the optimized formula, and provide new ideas and references for later researchers to evaluate the efficacy of the compound Chinese medicine formula after optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1A A schematic flow chart of the main steps of a method for optimizing a compound Chinese medicine prescription provided by an embodiment of the present invention;

[0038] Figure 1B A schematic flow chart of specific steps of a method for optimizing a compound Chinese medicine prescription provided by an embodiment of the present invention;

[0039] Figure 2A A schematic flow chart of the main steps of a method for evaluating the efficacy of an improved and original compound Chinese medicine formula;

[0040] Figure 2B A schematic flow chart of specific steps of a method for evaluating the efficacy of an improved compound Chinese medicine prescription according to this embodiment;

[0041] Figure 3 It is a schematic diagram of the structure of a deep neural network;

[0042] Figure 4A This is a comparison of target pathway enrichment between the original Xiaoyao Pills formula (left) and the improved formula (right);

[0043] Figure 4B This is a comparison chart of target pathway enrichment between the original (left) and improved (right) Erhuang Quzhi recipe;

[0044] Figure 4C This is a comparison chart of target pathway enrichment between the original formula of Qishen Yiqi Granule (left) and the improved formula (right);

[0045] Figure 5A Comparison of target tissue enrichment of Xiaoyao Pills original formula (left) and improved formula (right);

[0046] Figure 5B Comparison of target tissue enrichment of the original Erhuang Quzhi recipe (left) and the improved recipe (right);

[0047] Figure 5C Comparison of target tissue enrichment of the original formula of Qishen Yiqi Granule (left) and the improved formula (right);

[0048] Figure 6 To investigate the effects of Erhuang Quzhi recipe and its improved recipe on serum clarity, liver morphology, liver tissue pathology, and liver lipid droplet aggregation in hyperlipidemic mice.

[0049] Figure 7 This is a diagram showing the effect of Qishen Yiqi Dropping Pills and its modified formula on improving neurological deficits. (a) indicates that the neurological deficit score was tested using the modified Neurological Severity Score (mNSS) and Neurological Evaluation Scale (NES). Normal mice showed 0 points in mNSS and 15 points in NES. The horizontal axis represents different groups: the sham group is the sham operation group, the sham+T541-20 group is the sham operation + Qishen Yiqi Dropping Pills modified formula 20 mg / kg group, the Thrombus group is the thrombosis model group, the Thrombus+T541-20 group is the thrombosis model + Qishen Yiqi Dropping Pills modified formula 20 mg / kg group, the Thrombus+tPA group is the thrombosis model + tissue plasminogen activator group, and the Thrombus+tPA+T541 The -5 group was the thrombosis model + tissue plasminogen activator + improved Qishen Yiqi Dropping Pills 5 mg / kg group, the Thrombus+tPA+T541-10 group was the thrombosis model + tissue plasminogen activator + improved Qishen Yiqi Dropping Pills 10 mg / kg group, and the Thrombus+tPA+T541-20 group was the thrombosis model + tissue plasminogen activator + improved Qishen Yiqi Dropping Pills 20 mg / kg group. The vertical axis was the neurological deficit score. (b) represents the neurological deficit score, with the horizontal axis representing different groups: Sham group is the sham operation group, I / R group is the ischemia-reperfusion group, Eda group is the positive control edaravone group, QSYQ-L+I / R group is the low-dose Qishen Yiqi Dropping Pills + ischemia-reperfusion group, QSYQ-M+I / R group is the medium-dose Qishen Yiqi Dropping Pills + ischemia-reperfusion group, and QSYQ-H+I / R group is the high-dose Qishen Yiqi Dropping Pills + ischemia-reperfusion group. The vertical axis represents the neurological deficit score. DETAILED DESCRIPTION

[0050] In order to make those skilled in the art better understand the technical solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments in the present invention are only used to illustrate the present invention and do not limit the scope of the present invention in any way.

[0051] Figure 1A A flow chart of the main steps of a method for optimizing a compound Chinese medicine prescription provided by one embodiment of the present invention, the steps of the method comprising:

[0052] S1: Build a Chinese medicinal materials library and collect the chemical components of all Chinese medicinal materials in the library.

[0053] S2: Based on the chemical components of the Chinese medicinal materials collected in S1, an improved formula of the compound Chinese medicine is obtained based on computational analysis.

[0054] S3: Evaluate the efficacy of the improved prescription through calculation.

[0055] S4: Verify the efficacy of the improved formula through animal experiments.

[0056] Figure 1B A flow chart of the specific steps of a method for optimizing a compound Chinese medicine prescription of this embodiment includes the following steps:

[0057] S101. Collect Chinese medicinal materials through the Chinese Pharmacopoeia; collect the chemical components of Chinese medicinal materials through public databases.

[0058] S201. Based on the chemical components of Chinese medicinal materials collected in S101, a deep neural network (DNN) prediction model is constructed to judge the safety of adjuvants and adjuvants. If the safety is low, the corresponding medicinal materials are eliminated to obtain an improved prescription; if the safety is high, based on the ADME properties of the Chinese medicinal materials and the target network correlation analysis (specifically, the target set of Chinese medicinal materials collected through public databases and literature), after comprehensive calculation, the adjuvants or adjuvants ranked in the top 1-2 are eliminated to obtain an improved prescription.

[0059] S301. Specifically, the target set of Chinese medicinal materials is collected through public databases and literature (such as S201), and the gene set of diseases treated by compound Chinese medicine is obtained through public data. The random walk algorithm is used to calculate the correlation between the target set of the improved and original prescriptions and the disease gene set. Z-score>3 indicates that the drug and the disease are significantly correlated.

[0060] S401. Animal experiments verify the efficacy of the improved formula.

[0061] Figure 2A A flow chart of the main steps of a method for evaluating the efficacy of an improved compound Chinese medicine prescription provided by one embodiment of the present invention, the steps of the method comprising:

[0062] S5. Construct a chemical component library of compound Chinese medicine;

[0063] S6. Collect active targets of chemical components of compound Chinese medicine in S5 and conduct enrichment analysis of active targets;

[0064] S7. Calculate and evaluate the efficacy of the modified prescription;

[0065] S8. Animal experiments were used to verify the efficacy of the improved formula.

[0066] Figure 2B The flowchart of the specific steps of a method for evaluating the efficacy of an improved compound Chinese medicine prescription of this embodiment includes the following steps:

[0067] S501. Specifically, the Xingdou Cloud knowledge graph text mining tool is used to collect the experimentally verified chemical components of compound Chinese medicine.

[0068] S601. Specifically, based on the chemical components of the compound Chinese medicine in S401, the active targets of the chemical components of the original compound Chinese medicine and its improved formula are obtained through public databases, and the active targets are enriched and analyzed and compared respectively.

[0069] S701. Specifically, the genes of the diseases treated by the compound Chinese medicine are obtained through public data, and the network proximity of the active targets and disease genes of the original compound Chinese medicine and its improved formula is calculated to evaluate the efficacy of the improved formula on the disease. Z_closest<-3 means that the drug and the disease are significantly correlated.

[0070] S801. Animal experiments verify the efficacy of the improved formula.

[0071] Exemplarily, the specific implementation of step S201 is as follows:

[0072] The chemical structure molecular fingerprint ECFP6 (Extended-Connectivity Fingerprints 6) of the chemical components of Chinese medicinal materials is used to characterize the Chinese medicinal materials in the Chinese medicinal materials library in step S101, and a deep neural network (DNN) algorithm is used to build a Chinese medicinal materials safety prediction model. The model is trained and parameters are optimized through the training set, and the accuracy of the model is evaluated through the test set.

[0073] Deep neural networks such as Figure 3As shown in the figure, the bottom layer is the input layer, the middle layer is the hidden layer, and the last layer is the output layer. There are multiple hidden layers between the input layer and the output layer, and the layers are fully connected. The output of the input layer is represented by the vector χ, f(ω1χ+b1) is the output of the hidden layer, ω1 is the weight (connection coefficient), b1 is the bias, and the function f often uses the sigmoid function, relu function or tanh function; the output layer receives the output signal of the hidden layer and generates a prediction result, softmax(ω2χ1+b2) is the output of the output layer, χ1 is the output of the hidden layer f(ω1χ+b1), softmax converts the output of the neural network into a probability distribution That is, a Chinese herbal medicine (X) is classified into different safety categories. Here, the DNN model consists of 2 hidden layers, each with 50 neurons, relu as the activation function, adam algorithm as the optimization algorithm, cross entropy loss function, and the maximum number of iterations is 500. All Chinese medicinal materials are divided into training set and test set in a ratio of 8:2, and 5-fold cross validation is used for parameter optimization. The test set is used to evaluate the model, and the F1 score, accuracy and precision indicators are used to evaluate the performance of the prediction model. The DNN model predicts the test set results: macro-F1 score = 0.49, recall = 0.51, precision = 0.51, which is better than traditional machine learning models (random forest, support vector machine, nearest neighbor algorithm, extreme gradient boosting tree algorithm and naive Bayes algorithm).

[0074] The DNN model is used to predict the safety of adjuvants and adjuvant drugs respectively (definition of safety prediction: a prediction result of "0" indicates high safety; a prediction result of "1" indicates extremely low safety; a prediction result of "2" indicates low safety; a prediction result of "3" indicates relatively low safety).

[0075] ADME pharmacokinetic method (absorption, distribution, metabolism, excretion) is a very important method in drug design and drug screening. Based on DNN prediction of adjuvants and adjuvants with high safety, ADME prediction was performed on the chemical components of adjuvants and adjuvants, and the ADME property data set of the corresponding chemical components was obtained. Adjuvants and adjuvants were screened, eliminated and sorted according to different evaluation indicators, including intestinal absorption (weighted ratio: 0.4-0.5), oral availability (weighted ratio: 0.4-0.5) and Lipinski rule (weighted ratio: 0.2-0). The intestine is the main absorption site of drugs, and intestinal absorption of oral drugs is a necessary condition to show that they have significant therapeutic effects. Oral bioavailability reflects the proportion of the amount of the drug entering the human circulation. It describes the percentage of the oral drug absorbed by the gastrointestinal tract and the amount of the drug that reaches the systemic circulation through the liver. It is the most common pharmacokinetic parameter used for drug screening. Lipinski's Rule is an empirical rule for evaluating whether a compound can be used as a drug, or whether a compound with pharmacological or biological activity can be an oral drug. Lipinski's Rule means that a small molecule drug must have the following properties: 1) molecular weight less than 500; 2) number of hydrogen bond donors less than 5; 3) number of hydrogen bond acceptors less than 10; 4) lipid-water partition coefficient less than 5; 5) number of rotatable bonds not exceeding 10. Compounds that meet Lipinski's Rule will have better pharmacokinetic properties and higher bioavailability during metabolism in the body.

[0076] Target network correlation analysis: Degree centrality is the most direct metric to characterize target centrality in network analysis. The greater the target degree of a target, the higher the degree centrality of the target, and the more important the target is in the network. Closeness centrality uses the characteristics of the original square network. The more important the position of a target in the entire structure (the shortest distance between the target and other targets is very small), the higher the network centrality of the target. MCODE (Molecular Complex Detection) uses the relationship between edges and nodes to find key subnetworks and core targets in the network, and evaluate whether the target is located in the key network and whether it is a core gene.

[0077] Exemplarily, the specific implementation of step S301 is as follows:

[0078] The drug target set and the disease gene set are used as seed node sets to run the Random Walk with Restart (RWR) algorithm based on the given background network. In this way, the impact score vectors of the two sets of seed nodes on all nodes on the background network are obtained (representing the impact score of the seed gene on each node on the background network). Then the Pearson correlation coefficient of the two sets of score vectors is calculated, denoted as Cor, and finally the Zscore value is calculated by random sampling to evaluate the significance of the correlation between the drug target and the disease gene (Z score>3 means that the drug and the disease are correlated), as follows:

[0079]

[0080] Among them, E(Cor) and δ(Cor) represent the mean and standard deviation of the correlation coefficients calculated between drug targets and the same number of gene sets randomly selected from the background network for diseases (random sampling 1000 times). The RWR algorithm was performed using the dnet package, and the restart probability was set to 0.75 (default value).

[0081] Exemplarily, the specific implementation of step S501 is as follows:

[0082] The workflow of Xingdou Cloud Knowledge Graph Text Mining Tool mainly includes the following steps:

[0083] 1) Document download: Download the title and abstract information of the document from CNKI and Pubmed databases;

[0084] 2) Construct a local entity alias database: including five entity types: herbal medicine (Herb), chemical composition (Ingredient), disease (Disease), drug (Drug), and gene (Gene);

[0085] 3) Data query: Convert the format of the downloaded literature titles and abstract information and store them in the open source Elasticsearch database (https: / / www.elastic.co / cn / elasticsearch / , Elasticsearch is a distributed, highly scalable, and highly real-time search and data analysis engine. The user submits the data to the Elasticsearch database, and the corresponding sentence is segmented through the word segmentation controller, and its weight and word segmentation results are stored in the data. When the user searches for the data, the results are ranked and scored according to the weight, and finally the returned results are presented to the user); after entering the search for the compound Chinese medicine name (standard name and alias) on the Xingdou Cloud Knowledge Graph website (http: / / literature.tasly.com / self.php), the website background script program calls the entity library to determine the entity type of the compound Chinese medicine name and standardize the entity name. Search for relevant literature through the standard name and alias of the compound Chinese medicine to obtain literature information related to the compound Chinese medicine;

[0086] 4) Network construction: The literature information related to compound Chinese medicine obtained through the above steps is processed and a knowledge graph is constructed. In the knowledge graph, the name of the compound Chinese medicine is taken as the main node, and other entities that appear in the same document (such as chemical ingredients) are taken as child nodes. If two child nodes (such as chemical ingredients and genes) also exist in the same document, the two child nodes are considered to be related (that is, the two child nodes will be connected). In addition, the knowledge graph does not display all child nodes. Based on experience, the number of child nodes of different types is limited (that is, only nodes with a higher number of literature support are displayed). If you want to obtain all entities of a certain type, you need to obtain "Knowledge Graph->Associated (Entity Type) Information" on the search results page.

[0087] The Xingdou Cloud knowledge graph text mining tool was used to obtain the experimentally verified chemical components in the original formula of compound Chinese medicine and its improved formula. After standardizing the ingredient names and removing duplicate data, the corresponding chemical component library was obtained.

[0088] Exemplarily, the specific implementation of step S601 is as follows:

[0089] Enrichment analysis includes pathway enrichment analysis and tissue enrichment analysis:

[0090] Pathway enrichment analysis: Active targets are integrated together through specific algorithms to more intuitively display the relationship between active targets and signal pathways, and discover significant signal pathways related to the active target library.

[0091] Tissue enrichment analysis: Every cell in the human body has almost the same copy of the genome, but each tissue only expresses a subset of genes required to achieve its specific function. Therefore, the present invention also compares the enrichment expression of active targets of the chemical components of the original formula of the compound Chinese medicine and its improved formula in relevant tissues.

[0092] Exemplarily, the specific implementation of step S701 is as follows:

[0093] Disease-related genes are not randomly distributed on the background network, but tend to cluster to form local modules, called disease network modules. The effect of a drug on a disease can be evaluated by calculating the network proximity between the drug (X) and the disease (Y), defined as d(X,Y), as follows:

[0094]

[0095] Where d(x,y) represents the shortest path length between the drug target (x) and the disease gene (y), and ||X|| represents the number of drug targets. The significance of the drug-disease network proximity is evaluated by calculating the Z_closest value by random sampling. In the random sampling, the drug target is kept unchanged, and the randomly selected disease genes have the same number and similar degree distribution (on the background network) as the actual disease genes. The number of random samplings is 1000. The calculation formula is as follows:

[0096] Z_closest=(d-μ) / δ

[0097] Among them, μ and δ are the mean and standard deviation of the drug-disease network proximity calculated by random sampling. The smaller the value, the more significant the result (Z_closest<-3 means that the drug and disease are significantly correlated).

[0098] The experimental methods in the following examples are conventional methods unless otherwise specified. The medicinal materials, reagents, etc. used in the following examples are commercially available products unless otherwise specified.

[0099] Embodiment 1:

[0100] This example optimizes the original formula of Yangxueshengnao granules (original formula) based on the method of the present invention, and verifies the efficacy of the original formula (original formula) and the optimized formula (modified formula) in treating headache. The specific implementation steps are as follows:

[0101] Step 1: Build a database of Chinese medicinal materials and collect chemical components of Chinese medicinal materials

[0102] The Chinese herbal medicine database was constructed from the Pharmacopoeia of the People's Republic of China (2020 edition), which mainly includes 618 Chinese herbal medicines. The chemical components of the 618 Chinese herbal medicines are from the database shown in Table 1.

[0103] Table 1. Database

[0104]

[0105]

[0106] Step 2: Computational analysis to obtain improved solutions

[0107] Extended connectivity fingerprint was used to characterize 618 Chinese medicinal materials in the Chinese medicinal materials database based on compound structure characteristics, and a deep neural network (DNN) prediction model was constructed.

[0108] Yangxueshengnao granules are composed of 11 Chinese medicinal materials. The original formula includes white peony root, angelica root, Uncaria rhynchophylla, Millettia reticulata, Cassia seed, Chuanxiong, Rehmannia root, Asarum, Herba Prunellae, Corydalis yanhusuo and mother of pearl. Among them, the monarch drugs are angelica root and Chuanxiong; the minister drugs are white peony root, Uncaria rhynchophylla, Rehmannia root and mother of pearl; the adjuvant drugs are Millettia reticulata, Herba Prunellae, Corydalis yanhusuo and Cassia seed; and the messenger drug is Asarum. Referring to the compatibility principle of "monarch, minister, assistant and messenger", the Chinese medicinal materials of the monarch and minister drugs of Yangxueshengnao granules are retained, and the safety of the adjuvant and messenger drugs (i.e. Millettia reticulata, Herba Prunellae, Corydalis yanhusuo, Cassia seed and Asarum) are predicted by the DNN prediction model. The prediction results of the DNN model show that Asarum has low safety (DNN prediction result is "2"), while Millettia reticulata, Herba Prunellae, Cassia seed and Corydalis yanhusuo have high safety (DNN prediction results are all "0"). Therefore, Asarum is removed as an improved formula.

[0109] The above original prescription consists of: Angelica sinensis 253.5g, Ligusticum chuanxiong 253.5g, White Peony 202.7g, Rehmannia glutinosa 202.7g, Uncaria rhynchophylla 506.8g, Millettia spatholobi 506.8g, Prunella vulgaris 506.8g, Cassia seed 506.8g, Mother of pearl 506.8g, Corydalis 253.5g, and Asarum 50.5g.

[0110] The above improved prescription consists of: Angelica sinensis 253.5g, Ligusticum chuanxiong 253.5g, White Peony 202.7g, Rehmannia glutinosa 202.7g, Uncaria rhynchophylla 506.8g, Millettia spatholobi 506.8g, Prunella vulgaris 506.8g, Cassia seed 506.8g, Mother of pearl 506.8g, and Corydalis 253.5g.

[0111] Step 3: Calculation method to evaluate the efficacy of the modified prescription

[0112] To obtain the target set of Chinese medicinal materials in compound Chinese patent medicines: the 11 Chinese medicinal materials of Yangxueshengnao Granule (white peony root, Chinese angelica root, Uncaria rhynchophylla, Millettia reticulata, Cassia seed, Ligusticum chuanxiong, Rehmannia root, Asarum, Prunella vulgaris, Corydalis yanhusuo and mother-of-pearl) were input into the Chinese herbal medicine database LTM-TCM (https: / / testxdy.tasly.com / # / tcm / home) in Chinese or pinyin, and all the literature targets (experimentally verified targets) of each medicinal material were collected. After standardizing the target names and removing duplicate data, 245 targets were obtained for the original formula of Yangxueshengnao Granule and 238 targets were obtained for its improved formula, which were used as the target sets of the original formula and the improved formula, respectively.

[0113] Obtaining the primary headache gene set: Using "Primary headache" as the keyword, disease-related genes were collected in the following databases: GeneCards (https: / / www.genecards.org / ), OMIM (https: / / omim.org / ), Open Targets (https: / / www.targetvalidation.org / ), DisGeNET (https: / / www.disgenet.org / ), MalaCards (https: / / www.malacards.org / ), HGNC (https: / / www.genenames.org / ), Reactome (https: / / reactome.org / ), MetaCore (https: / / portal.genego.com / ) and KEGG (https: / / www.kegg.jp / ). After removing duplicate data and performing manual correction, 567 primary headache genes were obtained, which were used as the primary headache gene set.

[0114] As shown in Table 2, the correlation between the target set of the original and modified prescriptions and the primary headache gene set was studied based on the network propagation perspective, and the efficacy of the original and modified prescriptions on primary headache was evaluated. The results of the random walk algorithm calculation showed that Zscore>3, indicating that the modified prescription of Yangxueshengnao granules and primary headache had a good correlation.

[0115] Table 2. Correlation between the target set of the modified and original Yangxueqingnao granule and the primary headache gene set

[0116]

[0117] Step 4: Animal experiments to verify the efficacy of the improved formula

[0118] To observe the analgesic effects of the original and improved Yangxueqingnao granules on mice (Patent Publication No.: CN1919273A).

[0119] Blank control group: distilled water; positive control group: enteric-coated aspirin tablets.

[0120] Experimental group: original and improved formula of Yangxueqingnao granules.

[0121] Sample number: 112 mice in each of the two groups.

[0122] Efficacy observation indicators: Evaluation of the analgesic effect of the drug after acetic acid-induced pain in mice, and evaluation of the analgesic intensity of the drug after light radiation-induced pain in mice.

[0123] The results are shown in Table 3 and Table 4: Both the original and modified Yangxueqingnao granules can significantly inhibit the number of body twists caused by acid in mice, prolong the tail-flicking time caused by light irradiation in mice, and show different degrees of analgesic effect. This result is consistent with the prediction and calculation results obtained by the method of the present invention.

[0124] Table 3. Effects of the original and modified Yangxueshengnao granules on acetic acid-induced pain in mice

[0125] Group Dosage (g / kg) Number of animals 15min writhing inhibition rate (times) Inhibition rate % Control group - 14 29.36±7.78 - YXQN-01 28 14 16.50±7.56*** 49.88 YXQN-01 14 14 18.93±6.79** 45.99 YXQN-01 7 14 22.36±9.48* 33.58 YXQN-02 28 14 17.71±9.56** 44.53 YXQN-02 14 14 13.93±7.14*** 62.29 YXQN-02 7 14 21.71±10.03* 43.07 Aspirin enteric-coated tablets 0.2 14 8.57±4.85*** 69.83

[0126] Note: YXQN: Yangxueshengnao granules; compared with the control group, *P<0.05, **P<0.01, ***P<0.001.

[0127] Table 4. Effects of the original and modified Yangxueshengnao granules on thermal radiation tail-flick-induced pain in mice

[0128]

[0129]

[0130] Note: YXQN: Yangxueshengnao granules; compared with the control group, *P<0.05, **P<0.01.

[0131] In summary, computational analysis and animal experimental verification showed that the efficacy of the improved Yangxueqingnao granule after optimization of this method in the treatment of primary headache is equivalent to that of the original Yangxueqingnao granule.

[0132] Embodiment 2:

[0133] This example optimizes the original formula of Xiaoyao Pills based on the method of the present invention, and verifies the efficacy of the original formula (original formula) and the optimized formula (modified formula) in treating depression. The specific implementation steps are as follows:

[0134] Step 1: Collect chemical components of Chinese medicinal materials

[0135] Xiaoyao Pills are composed of 7 Chinese medicinal materials, including Bupleurum, Angelica, White Peony, Atractylodes, Poria, Licorice and Menthol. Among them, the main drug is Bupleurum; the auxiliary drugs are White Peony and Angelica; the adjuvant drugs are Atractylodes, Poria and Licorice; and the guiding drug is Menthol. The 7 Chinese medicinal materials of Xiaoyao Pills are respectively input into the Chinese medicinal material database in Chinese or Pinyin (the database is detailed in Example 1: Step 1), and all the chemical components of each medicinal material are collected. After merging and sorting and removing duplicate data, 2037 chemical components are finally obtained.

[0136] Step 2: Computational analysis to obtain improved solutions

[0137] Referring to the compatibility principle of "monarch, minister, assistant and messenger", the Chinese medicinal materials of the monarch and minister of Xiaoyao Pills were retained, and the safety of the assistant and messenger (i.e., Atractylodes macrocephala, Poria cocos, Licorice and Menthol) was predicted by the DNN prediction model constructed in Example 1. The prediction results of the DNN model all showed that the safety of the assistant and messenger was high (the DNN prediction results were all "0").

[0138] ADME analysis: After comparing the chemical component sets of Atractylodes macrocephala, Poria cocos, Licorice and Mint with the chemical component sets of the main and auxiliary drugs (Bupleurum, Angelica sinensis and White Peony Root) one by one, the number of unique components of Licorice (359) is much higher than that of Mint (206), Atractylodes macrocephala (124) and Poria cocos (99). Licorice is retained because it may play a unique pharmacodynamic role. Atractylodes macrocephala, Poria cocos and Mint are eliminated, evaluated and ranked according to different ADME properties. The evaluation indicators include intestinal absorption, oral availability and Lipinski rule. In the present invention, the screening threshold of intestinal absorption and oral availability of Chinese herbal medicine chemical components is set to be greater than 0.7 (0.7 represents the probability of intestinal absorption and oral availability being less than 30%, that is, the chemical components with low intestinal absorption and oral availability in Chinese herbal medicine are eliminated). The purpose of setting the threshold is: 1) to make the chemical components obtained by screening have as high intestinal absorption rate and oral availability as possible; 2) the chemical components obtained by screening can represent the pharmacological data properties of the entire Chinese medicine compound as comprehensively as possible.

[0139] 1. Intestinal absorption: The intestinal absorption values ​​in the ADME property data set of the corresponding chemical components of Atractylodes macrocephala, Poria cocos and mint were screened according to the set thresholds, and the components with intestinal absorption greater than 0.7 were eliminated. Finally, the ranking was obtained according to the number of eliminated components of Chinese medicinal materials: Poria cocos (37) > Mint (21) > Atractylodes macrocephala (12).

[0140] 2. Oral availability: The oral availability values ​​in the ADME property data set of the corresponding chemical components of Atractylodes macrocephala, Poria cocos and mint were screened according to the set thresholds, and the components with oral availability greater than 0.7 were eliminated. Finally, the ranking was obtained according to the number of eliminated components of Chinese medicinal materials: Mint (121) > Atractylodes macrocephala (86) > Poria cocos (64).

[0141] 3. Lipinski rule: The Lipinski rule in the ADME property data set of the chemical components corresponding to Atractylodes macrocephala, Poria cocos and Mint was screened, and the chemical components that did not conform to the Lipinski rule were eliminated (Rejected). Finally, the number of rejected components of Chinese medicinal materials was ranked as follows: Poria cocos (41) > Mint (9) > Atractylodes macrocephala (5).

[0142] Based on the above steps, the contribution of the present invention is assigned a weighted score according to different evaluation indicators and the screening and ranking of Chinese medicinal materials (intestinal absorption = oral availability and > Lipinski rule), that is, the more important the evaluation indicator is, the greater the value assigned to the evaluation indicator; the higher the ranking of the Chinese medicinal material screening, the greater the value assigned to the Chinese medicinal material. In view of this, the weighted ratio of intestinal absorption, oral availability and Lipinski rule in the present invention is 0.4:0.4:0.2, the weighted ratio of the Chinese medicinal material screening and ranking score is 0.3:0.2:0.1, and the weighted score of the Chinese medicinal material score = (indicator 1 + indicator 2 + indicator 3) * Chinese medicinal material screening and ranking score. Therefore, the ranking of Chinese medicinal materials that need to be eliminated is obtained: mint (score 0.24) > Poria cocos (score 0.22) > Atractylodes macrocephala (score 0.14). The higher the ranking, the more it needs to be eliminated.

[0143] Target network correlation analysis: The seven medicinal materials of Xiaoyao Pills were input into the Xingdouyun Knowledge Graph Library (https: / / literature.tasly.com / ) in Chinese or pinyin, and all the literature targets (experimentally verified targets) of each medicinal material were collected. After standardizing the target names and removing duplicate data, 384 targets were finally obtained. In order to avoid false positive results, this step gave priority to targets with a literature support number greater than 1 (112), that is, to obtain the target set corresponding to the original prescription, and the original prescription target set was input into the STRING database (https: / / string-db.org / ) to obtain the original prescription target network. At the same time, all targets of the monarch and ministerial drugs were retained, and Atractylodes macrocephala, Poria cocos, and mint were randomly removed. As shown in Tables 5 and 6, Improved Prescription 1 (reduced Atractylodes macrocephala), Improved Prescription 2 (reduced Poria cocos), Improved Prescription 3 (reduced Mint) and their corresponding target sets were obtained, respectively.

[0144] Table 5. Composition of the original and improved Xiaoyao Pills

[0145] composition original party Bupleurum, Chinese Angelica, White Peony Root, Atractylodes Macrocephala, Poria, Licorice, Mint Improved recipe 1 (reducing white atractylodes) Bupleurum, Chinese Angelica, White Peony Root, Poria, Licorice, Mint Improved recipe 2 (reducing Poria cocos) Bupleurum, Chinese Angelica, White Peony Root, Atractylodes Macrocephala, Licorice, Mint Improved recipe 2 (minus mint) Bupleurum, Chinese Angelica, White Peony Root, Atractylodes Macrocephala, Poria, Licorice

[0146] Table 6. Target set of the original and improved Xiaoyao Pills

[0147]

[0148]

[0149] Screening and sorting are performed based on the centrality and importance of the target. The detailed steps are as follows:

[0150] 1. Degree centrality: Based on the above steps, the targets of Atractylodes macrocephala, Poria cocos and Mentha were randomly removed to obtain the target set of Improved Recipe 1, the target set of Improved Recipe 2 and the target set of Improved Recipe 3. There was no difference between the target set of Improved Recipe 3 and the target set of the original recipe (Table 6), indicating that Improved Recipe 3 (minus Mentha) had the highest similarity with the original recipe, and the degree centrality ranking was: Improved Recipe 3 < Improved Recipe 2 < Improved Recipe 1, so based on degree centrality, Improved Recipe 3 was preferred (Table 7).

[0151] 2. Target centrality: Based on the above steps, the original target set was input into the STRING database to obtain the original target network, and the network diagram was drawn using Cytoscape software. The centrality of the different targets between the improved and original recipes in the network was compared using Analyze Network, as shown in Table 7. The ranking according to the centrality evaluation was: Improved recipe 3 < Improved recipe 2 < Improved recipe 1. Therefore, based on the target centrality, Improved recipe 3 was preferred.

[0152] 3. Core targets: Based on the above steps, the original target set was input into the STRING database to obtain the original target network, the network diagram was drawn using Cytoscape software, and the scores of the different targets of the improved and original formulas in the network were compared by MCODE analysis. As shown in Table 7, according to the MCODE score ranking: Improved Formula 3 < Improved Formula 2 < Improved Formula 1, so based on the comparison of core targets, Improved Formula 3 is preferred.

[0153] Table 7. Importance of different targets between the original and modified Xiaoyao Pills

[0154]

[0155] Based on the above steps, weighted scores are assigned according to different indicators, that is, the more important the evaluation indicator is, the greater the value assigned to the evaluation indicator; the smaller the ranking of the improved target importance evaluation indicator, the greater the value assigned to the improved formula. In view of this, the weighted ratio of target similarity, centrality and MCODE scores is 0.4:0.3:0.3, and the weighted ratio of the improved formula ranking score is 0.3:0.2:0.1 (if the ranking is the same, the same value is assigned). The weighted score of the improved formula = (indicator 1 + indicator 2 + indicator 3) * improved formula ranking score. Therefore, the ranking of Chinese medicinal materials that need to be eliminated is obtained: mint (score 0.3) > Poria (score 0.2) > Atractylodes (score 0.1). The higher the ranking, the more it needs to be eliminated.

[0156] Based on the above ingredients and targets (1:1) screening, the total scores of adjuvants and adjuvants were: mint (Total score 0.54) > Poria (Total score 0.44) > Atractylodes (Total score 0.24). Mint and Poria were removed from the final Xiaoyao Pill formula (7 Chinese medicinal materials). Therefore, the optimized formula includes 5 Chinese medicinal materials, namely Bupleurum, Angelica, White Peony, Atractylodes and Licorice.

[0157] Step 3: Calculation method to evaluate the efficacy of the modified prescription

[0158] Using "Depression" as the keyword, we collected disease-related genes from the following databases: GeneCards (https: / / www.genecards.org / ), Open Targets (https: / / www.targetvalidation.org / ), OMIM (https: / / omim.org / ), DisGeNET (https: / / www.disgenet.org / ), GWAS (https: / / www.ebi.ac.uk / gwas / ), MalaCards (https: / / www.malacards.org / ), HGMD (https: / / www.hgmd.cf.ac.uk / ac / index.php), Reactome (https: / / reactome.org / ), MetaCore (https: / / portal.genego.com / ) and KEGG (https: / / www.kegg.jp / ). After removing duplicate data and performing manual correction, we obtained 272 depression genes, which were used as the depression gene set.

[0159] Based on the network communication perspective, the correlation between the target set of the original and improved Xiaoyao Pills and the depression gene set was studied to evaluate the efficacy of the original and improved prescriptions on depression. The results of the random walk algorithm showed that the improved Xiaoyao Pills were significantly correlated with depression (Z score>3), as shown in Table 8.

[0160] Table 8. Correlation between target sets and disease gene sets of the modified and original Xiaoyao Pills

[0161]

[0162] Step 4: Animal experiments to verify the efficacy of the improved formula

[0163] Compare the antidepressant effects of the original Xiaoyao Pills and its improved formula (Patent Publication No.: CN 114053337 A).

[0164] Blank control group: distilled water; positive control group: fluoxetine hydrochloride (commonly used to treat depression)

[0165] Experimental group: original and improved Xiaoyao Pills.

[0166] Sample number: 13 rats in each of the two groups.

[0167] Efficacy observation indicators: Body weight change, percentage of sugar water consumption, and open-box exercise score are the main behavioral evaluation indicators, which can reflect the symptoms of clinical depression patients such as loss of appetite, anhedonia, autonomous activity ability, and reduced curiosity about new environments; the antidepressant effect is evaluated by measuring the content of biochemical indicators (corticosterone CORT, dopamine DA, norepinephrine NE, 5-hydroxytryptamine 5-HT, brain-derived neurotrophic factor BDNF).

[0168] The results of efficacy indexes are shown in Table 9. The experimental results show that the effect of the modified prescription group on improving the symptoms of anorexia, weight loss, anhedonia, low activity and reduced curiosity about the new environment in depression model rats is second only to the fluoxetine hydrochloride group, and is better than the original prescription group. The detection of the content of biochemical indicators (CORT, DA, NE, 5-HT, BDNF) shows that the antidepressant effect of the modified prescription group is equivalent to that of the fluoxetine hydrochloride group, and is better than the original prescription.

[0169] Table 9. Antidepressant effects of the original and modified Xiaoyao Pills

[0170] Efficacy index Order of size of groups Weight gain Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group Sugar water consumption percentage Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group Horizontal Movement Score Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group Vertical Movement Score Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group CORT content in plasma Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group DA, NE, and 5-HT content in the cortex Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group BDNF content in the cortex Improved prescription group > Fluoxetine hydrochloride group > Xiaoyao Pills original prescription group Contents of DA, NE and 5-HT in hippocampus Fluoxetine hydrochloride group > modified formula group > Xiaoyao pill original formula group BDNF content in hippocampus Improved prescription group > Fluoxetine hydrochloride group > Xiaoyao Pills original prescription group total Fluoxetine hydrochloride group, modified prescription group > Xiaoyao Pills original prescription group

[0171] In summary, computational analysis showed that the efficacy of the improved Xiaoyao Pills optimized by this method in the treatment of primary headaches was equivalent to that of the original Xiaoyao Pills. Animal experiments showed that the antidepressant effect of the improved Xiaoyao Pills was equivalent to that of the original Xiaoyao Pills, proving that this optimization method is effective.

[0172] Embodiment three:

[0173] This example compares and analyzes the active targets of the chemical components of the original Xiaoyao Pills and the modified Xiaoyao Pills based on the method of the present invention, and evaluates the efficacy of the modified Xiaoyao Pills by calculating the network proximity between the original Xiaoyao Pills and the modified Xiaoyao Pills and depression. The specific implementation steps are as follows:

[0174] Step 1: Construct the chemical component library of Xiaoyao Pills

[0175] Xiaoyao Pills are composed of 7 Chinese medicinal materials, including Bupleurum, Angelica, White Peony, Atractylodes, Poria, Licorice and Mint. Its improved formula includes 5 Chinese medicinal materials, Bupleurum, Angelica, White Peony, Atractylodes and Licorice. All chemical components corresponding to Xiaoyao Pills were collected in the Xingdouyun Knowledge Graph Library using "Xiaoyao Pills", "xiao yao wan" and "Xiaoyao Pill" as keywords respectively. After standardizing the names of the ingredients and eliminating duplicate data, the original formula of Xiaoyao Pills obtained 18 chemical components, and its improved formula obtained 18 chemical components.

[0176] The chemical components contained in the original and improved formulas of Xiaoyao Pills are shown in Table 10.

[0177] Table 10. Chemical components of the original and improved Xiaoyao Pills

[0178]

[0179]

[0180] Step 2: Obtain active targets of Xiaoyao Pills chemical components and conduct enrichment analysis

[0181] Using the names of chemical components of Xiaoyao Pills as keywords, the related active targets of the chemical components of Xiaoyao Pills (18) were collected one by one in the PubChem (https: / / pubchem.ncbi.nlm.nih.gov / ), ChEMBL (https: / / www.ebi.ac.uk / chembl / ) and HERB (http: / / herb.ac.cn / ) databases. After removing duplicate targets, 292 active targets were obtained, and the improved formula obtained 292 active targets.

[0182] The active target libraries of the chemical components of the original and improved Xiaoyao Pills were entered into the DAVID database (https: / / david.ncifcrf.gov / home.jsp), and the species was selected as Homo sapiens for KEGG Pathway analysis. Figure 4A As shown, the active targets related to the original and improved Xiaoyao Pills were significantly enriched in fat synthesis, cancer-related and AGE-RAGE pathways, and there was no difference in the pathway enrichment of the active targets related to the improved and original formulas.

[0183] Tissue-specific expression analysis was performed using the TSEA tool (http: / / genetics.wustl.edu / jdlab / tsea / ), and the active target libraries of the chemical components of the original and improved Xiaoyao Pills were input, and the species was selected as Human for TSEA analysis. Figure 5AAs shown in the figure, the active targets related to the original and improved formulas of Xiaoyao Pills were significantly enriched in tissues such as blood, liver and colon; there was no difference in the tissue enrichment of the active targets related to the improved formula and the original formula.

[0184] Step 3: Evaluate the efficacy of the modified Xiaoyao Pills formula by calculation

[0185] Obtaining depression genes: Using "Depression" as the keyword, 272 depression genes were obtained (see Example 2 for details).

[0186] Based on the network proximity between drugs and diseases, the efficacy of the original and modified Xiaoyao Pills on depression was evaluated. The calculation results showed that the network proximity between the modified Xiaoyao Pills and depression was no different from that of the original prescription, and the modified Xiaoyao Pills was significantly correlated with depression (Z_closest<-3), as shown in Table 11.

[0187] Table 11. Network proximity between the original and modified Xiaoyao Pills and depression

[0188]

[0189] Step 4: Animal experiments to verify the efficacy of the modified Xiaoyao Pills formula

[0190] The experimental results show that the improved prescription group is better than the original prescription group in improving the symptoms of loss of appetite, weight loss, anhedonia, low activity ability and reduced curiosity about new environments in depression model rats, and the improved prescription group has a better antidepressant effect than the original prescription (see Example 2 for details).

[0191] In summary, after target enrichment analysis, network proximity calculation and animal experiment verification of the original and improved Xiaoyao Pills, it was shown that there was no difference in the efficacy between the improved Xiaoyao Pills and the original formula, proving that this method can be used to evaluate the efficacy of improved compound Chinese medicine formulas.

[0192] Embodiment 4:

[0193] This example compares and analyzes the active targets of the chemical components of the original and improved Erhuang Quzhi Recipe based on the method of the present invention, and evaluates the efficacy of the improved Erhuang Quzhi Recipe by calculating the network proximity between the original and improved Erhuang Quzhi Recipe and hyperlipidemia. The specific implementation steps are as follows:

[0194] Step 1: Construct the chemical component library of Erhuang Quzhi Recipe

[0195] Erhuang Quzhi recipe is composed of 13 Chinese medicinal materials including licorice, salvia miltiorrhiza, lotus leaf, rhubarb, turmeric, knotweed, kudzu root, atractylodes, oriental rhizome, gynostemma pentaphyllum, bombyx batryticatus, leech and qinglun stone, and its improved recipe includes 4 Chinese medicinal materials including knotweed, salvia miltiorrhiza, lotus leaf and turmeric. All chemical components corresponding to Erhuang Quzhi recipe were collected in Xingdouyun knowledge graph library with "Er huang qu zhi recipe" and "Er huang qu zhi" as keywords respectively. After standardizing the names of ingredients, eliminating duplicate data and manual correction, 29 chemical components were obtained in the original recipe of Erhuang Quzhi recipe, and 20 chemical components were obtained in its improved recipe.

[0196] The chemical components contained in the original and improved formula of Erhuang Quzhi Recipe are shown in Table 12.

[0197] Table 12. Chemical components of the original and improved Erhuang Quzhi recipe

[0198]

[0199] Step 2: Obtain active targets of chemical components of Erhuang Quzhi Recipe and conduct enrichment analysis

[0200] Using the chemical constituents of Erhuang Quzhi Recipe as keywords, the active targets of the chemical constituents of Erhuang Quzhi Recipe were collected one by one in the PubChem (https: / / pubchem.ncbi.nlm.nih.gov / ), ChEMBL (https: / / www.ebi.ac.uk / chembl / ) and HERB (http: / / herb.ac.cn / ) databases. After manual correction and removal of duplicate targets, 322 active targets were obtained for Erhuang Quzhi Recipe, and 306 active targets were obtained for its improved recipe.

[0201] The active target libraries of the chemical components of the original and improved Erhuang Quzhi recipe were entered into the DAVID database (https: / / david.ncifcrf.gov / home.jsp), and the species was selected as Homo sapiens for KEGG Pathway analysis. Figure 4B As shown, the active targets related to the original and improved Erhuang Quzhi recipe were significantly enriched in pathways related to fat synthesis and cancer, and there was no difference in the pathway enrichment of the active targets related to the improved and original recipes.

[0202] Tissue-specific expression analysis was performed using the TSEA tool (http: / / genetics.wustl.edu / jdlab / tsea / ), and the active target libraries of the chemical components of the original and improved Erhuang Quzhi recipe were input respectively, and the species was selected as Human for TSEA analysis. Figure 5BAs shown, the active targets related to the original and improved Erhuang Quzhi recipe were significantly enriched in tissues such as liver, blood and colon; there was no difference in tissue enrichment of the active targets related to the improved and original recipes.

[0203] Step 3: Evaluate the efficacy of the modified Erhuang Quzhi recipe by calculation method

[0204] Obtaining genes for hyperlipidemia: Using "Hyperlipidemia" as the keyword, disease-related genes were collected in the following databases: GeneCards (https: / / www.genecards.org / ), OMIM (https: / / omim.org / ), OpenTargets (https: / / www.targetvalidation.org / ), DisGeNET (https: / / www.disgenet.org / ), GWAS (https: / / www.ebi.ac.uk / gwas / ), MalaCards (https: / / www.malacards.org / ), HGM (https: / / www.hgmd.cf.ac.uk / ac / index.php), Reactome (https: / / reactome.org / ), MetaCore (https: / / portal.genego.com / ), and KEGG (https: / / www.kegg.jp / ). They were used as disease-related gene libraries and manually corrected, and 296 hyperlipidemia genes were obtained.

[0205] Based on the network proximity between drugs and diseases, the efficacy of the original and modified Erhuang Quzhi Recipe on hyperlipidemia was evaluated. The calculation results showed that the network proximity between the modified Erhuang Quzhi Recipe and hyperlipidemia was no different from that of the original recipe, and the modified Erhuang Quzhi Recipe was significantly correlated with hyperlipidemia (Z_closest<-3), as shown in Table 13.

[0206] Table 13. Network proximity between the original and modified Erhuang Quzhi recipe and hyperlipidemia

[0207]

[0208] Step 4: Animal experiment to verify the efficacy of the modified Erhuang Quzhi recipe

[0209] To compare the effects of the original Erhuang Quzhi recipe and its improved recipe in treating hyperlipidemia.

[0210] Blank control group: distilled water; hyperlipidemia group: distilled water; positive control group: bezafibrate (for lowering blood lipids).

[0211] Experimental group: original and improved Erhuang Quzhi recipe.

[0212] Sample number: 6 mice in each group. ICR mice were randomly divided into CON group, HPL group, EHQZF group (2.4g / kg), QZF-I group (2.4g / kg), QZF-II group (2.4g / kg) and BZBT group (60.67mg / kg). Except for CON group and HPL group, which were given purified water, the other groups were given corresponding drugs by gavage, once in the morning and once in the evening, for 13 days. On the 13th day, 1 hour after the last administration, the other groups, except CON group, were intraperitoneally injected with 500mg / kg of Tyloxapol, and the mice were forbidden to eat. After 14 h of food deprivation, the mice were euthanized, and serum was collected to detect the levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), malondialdehyde (MDA), and superoxide dismutase (SOD). Liver tissue homogenate was collected to detect the levels of interleukins IL-6 and IL-1β. Pathological observation was performed using hematoxylin-eosin (HE) staining and Oil Red O staining.

[0213] Efficacy observation indicators: Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), malondialdehyde (MDA), superoxide dismutase (SOD) and interleukin (IL)-6 and IL-1β in liver tissue were detected, and pathological observation was performed using hematoxylin-eosin (HE) staining and Oil Red O staining.

[0214] The experimental results show that both the original and modified Erhuang Quzhi recipe can significantly reduce serum ALT, AST, TG, TC, and LDL-C levels, inhibit MDA levels, and increase HDL-C and SOD levels; both the original and modified Erhuang Quzhi recipe can reduce the content of IL-6 and IL-1β in liver tissue, and the modified recipe has better effect. At the same time, both the original and modified Erhuang Quzhi recipe can improve the serum clarity and liver pathological state of mice, reduce liver inflammatory cell infiltration and lipid droplet accumulation, and the modified recipe has a significant effect. It is proved that the modified recipe can treat hyperlipidemia through anti-oxidation, improving liver damage, reducing liver inflammation and lipid droplet accumulation. Figure 6 shown.

[0215] In summary, after target enrichment analysis, network proximity calculation and animal experiment verification of the original and improved Erhuang Quzhi recipe, it was shown that there was no significant difference in the efficacy between the improved and original recipes of Erhuang Quzhi recipe, proving that this method can be used to evaluate the efficacy of improved compound Chinese medicine recipes.

[0216] Embodiment five:

[0217] This example compares and analyzes the active targets of the chemical components of the original and improved formulas of Qishen Yiqi Dropping Pills based on the method of the present invention, and evaluates the efficacy of the improved formula of Qishen Yiqi Dropping Pills by calculating the network proximity between the original and improved formulas of Qishen Yiqi Dropping Pills and ischemic stroke. The specific implementation steps are as follows:

[0218] Step 1: Construct the chemical component library of Qishen Yiqi Dropping Pills

[0219] Qishen Yiqi Dripping Pills is composed of four Chinese medicinal materials: Astragalus, Salvia miltiorrhiza, Panax notoginseng and Dalbergia odorifera. Its improved formula includes three Chinese medicinal materials: Astragalus, Salvia miltiorrhiza and Panax notoginseng. All chemical components corresponding to Qishen Yiqi Dripping Pills were collected in the Xingdouyun Knowledge Graph Library using "Qishen Yiqi Dripping Pills" and "Qishen Yiqi Dripping Pills" as keywords, respectively. After standardizing the names of ingredients, eliminating duplicate data and manual correction, 26 chemical components were obtained from the original formula of Qishen Yiqi Dripping Pills, and 23 chemical components were obtained from its improved formula.

[0220] The chemical components contained in the original and improved formulas of Qishen Yiqi Dropping Pills are shown in Table 14.

[0221] Table 14. Chemical components of the original and improved formulas of Qishen Yiqi Dropping Pills

[0222]

[0223] Step 2: Obtain active targets of chemical components of Qishen Yiqi Dropping Pills and conduct enrichment analysis

[0224] Using the chemical constituents of Qishen Yiqi Dropping Pills as keywords, the active targets of the chemical constituents of Qishen Yiqi Dropping Pills were collected one by one in the PubChem (https: / / pubchem.ncbi.nlm.nih.gov / ), ChEMBL (https: / / www.ebi.ac.uk / chembl / ) and HERB (http: / / herb.ac.cn / ) databases. After manual correction and elimination of duplicate targets, 315 active targets were obtained for Qishen Yiqi Dropping Pills, and 314 active targets were obtained for its improved formula.

[0225] The active target libraries of the chemical components of the original and improved formulas of Qishen Yiqi Dropping Pills were entered into the DAVID database (https: / / david.ncifcrf.gov / home.jsp), and the species was selected as Homo sapiens for KEGG Pathway analysis. Figure 4C As shown in the figure, the active targets related to the original and improved formulas of Qishen Yiqi Dropping Pills were significantly enriched in the AGE-RAGE pathway and cancer-related pathways, and there was no difference in the pathway enrichment of the active targets related to the improved formula and the original formula.

[0226] Tissue-specific expression analysis was performed using the TSEA tool (http: / / genetics.wustl.edu / jdlab / tsea / ), and the active target libraries of the chemical components of the original and improved formulas of Qishen Yiqi Dropping Pills were input, and the species was selected as Human for TSEA analysis. Figure 5C As shown, the active targets related to the original and improved formulas of Qishen Yiqi Dropping Pills were significantly enriched in tissues such as heart, liver, lung and colon; there was no difference in tissue enrichment of the active targets related to the improved formula and the original formula.

[0227] Step 3: Evaluate the efficacy of the modified formula of Qishen Yiqi Dropping Pills by calculation method

[0228] Obtaining genes for ischemic stroke: Using "Ischemic stroke" as the keyword, disease-related genes were collected in the following databases: GeneCards (https: / / www.genecards.org / ), OMIM (https: / / omim.org / ), Open Targets (https: / / www.targetvalidation.org / ), DisGeNET (https: / / www.disgenet.org / ), GWAS (https: / / www.ebi.ac.uk / gwas / ), MalaCards (https: / / www.malacards.org / ), HGM (https: / / www.hgmd.cf.ac.uk / ac / index.php), Reactome (https: / / reactome.org / ), MetaCore (https: / / portal.genego.com / ) and KEGG (https: / / www.kegg.jp / ). They were used as disease-related gene libraries and manually corrected to obtain 312 ischemic stroke genes.

[0229] Based on the network proximity between drugs and diseases, the efficacy of the original and modified Qishen Yiqi Dropping Pills on diseases was evaluated. The calculation results showed that there was no significant difference between the network proximity between the modified Qishen Yiqi Dropping Pills and ischemic stroke and the original prescription, and the modified Qishen Yiqi Dropping Pills was significantly correlated with ischemic stroke (Z_closest<-3), as shown in Table 15.

[0230] Table 15. Network proximity between the original and improved formulas of Qishen Yiqi Dropping Pills and diseases

[0231]

[0232] Step 4: Animal experiment to verify the efficacy of the modified formula of Qishen Yiqi Dropping Pills

[0233] To compare the effects of Qishen Yiqi Dropping Pills original formula and its improved formula in the treatment of ischemic stroke.

[0234] Control groups: sham operation group, model group, positive control group.

[0235] Experimental group: Improved formula of Qishen Yiqi Dropping Pills.

[0236] Efficacy observation indicators: The protective effect of Qishen Yiqi Dropping Pills on cerebral ischemia / reperfusion injury was evaluated by assessing neurological deficits, cerebral infarction, cerebral edema and blood-brain barrier function.

[0237] Experimental methods:

[0238] Carotid artery thrombosis model and animal treatment

[0239] Male C57BL / 6N mice (21±2 g) were obtained from the Animal Center of Peking University Science Center (Beijing, Certificate No. SCXK2006-0008). Carotid artery thrombosis model was established: animals were anesthetized by intraperitoneal injection of sodium pentobarbital (2%; 45 mg / kg), the right common carotid artery was isolated, and thrombus formation was induced by wrapping with 10% FeCl3-soaked 1 mm wide filter paper for 3 minutes, and the sham operation group was replaced with normal saline. Carotid artery thrombosis was detected over time, and only animals with maximum thrombus occlusion of >80% of arterial diameter 4.5 hours after ischemia were used for further experiments. The animals were then randomly divided into different groups and treated with tissue plasminogen activator tPA (10 mg / kg; Boehringer Ingelheim, Germany) with or without the modified formula of Qishen Yiqi Drop Pills (5, 10, or 20 mg / kg).

[0240] Qishen Yiqi dripping pills (QSYQ) (NMPN Z20030139) were provided by Tasly Pharmaceutical Group Co., Ltd. (Tianjin, China). Qishen Yiqi dripping pills were dissolved in ultrapure water to obtain solutions with concentrations of 5.69 mg / mL, 11.38 mg / mL, and 22.75 mg / mL for subsequent experiments. Male 6-week-old ICR mice (22±2 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, Certificate No.: SCXK Jing 2014-0013). The experimental mice were divided into sham operation group (Sham group), ischemia-reperfusion group (I / R group), positive control group (edaravone, Eda group), QSYQ low dose (5.69 mg / mL) + I / R group, medium dose (11.38 mg / mL) + I / R group, and high dose (22.75 mg / mL) + I / R group. The mice in the treatment group were gavaged with QSYQ once a day for 7 days before the ischemia-reperfusion surgery. The positive control drug edaravone (2 mg / mL, drug approval number: H20031342, purchased from Nanjing Xiansheng Dongyuan Pharmaceutical Co., Ltd., Nanjing) was intravenously injected once 50 minutes after ischemia. The rats in the sham operation group and the I / R model group were gavaged with 0.9% saline (10 mL / kg) for 7 consecutive days.

[0241] Neurological deficits are assessed in two ways:

[0242] (1) Neurological deficits were scored using the modified neurological severity score (mNSS) and the 15-point neurological evaluation scale (NES). Normal mice scored 0 in the mNSS and 15 in the NES. Figure 7 As shown in (a) in .

[0243] (2) Neurological deficits of experimental animals were examined 24 hours after reperfusion. Score 0, no neurological deficit (normal and spontaneous movement); score 1, mild focal neurological deficit (failure to fully extend the contralateral forelimb); score 2, moderate focal neurological deficit (repeatedly circling to the contralateral side); score 3, severe focal neurological deficit (falling to the contralateral side); score 4, low consciousness or death, unable to walk independently. Figure 7 As shown in (b) in .

[0244] The experimental results showed that the original and modified formulas of Qishen Yiqi Dropping Pills significantly reduced the neurological deficit score, relieved the related vascular edema and bleeding, and reduced the damage of the blood-brain barrier. It proved that the original and modified formulas of Qishen Yiqi Dropping Pills have a protective effect on cerebral ischemia / reperfusion injury. Figure 7 shown.

[0245] In summary, after target enrichment analysis, network proximity calculation and animal experimental verification of the original and improved formulas of Qishen Yiqi Dropping Pills, it was shown that there was no significant difference in the efficacy between the improved formula of Qishen Yiqi Dropping Pills and the original formula, and both had a protective effect on cerebral ischemia / reperfusion injury, proving that this method can be used to evaluate the efficacy of improved formulas of compound Chinese medicines.

[0246] The specific embodiments of the present invention disclosed above are intended to help understand the content of the present invention and implement it accordingly. It can be understood by those skilled in the art that various replacements, 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 contents disclosed in the embodiments of this specification, and the scope of protection of the present invention shall be subject to the scope defined in the claims.

Claims

1. A method for optimizing a compound Chinese medicine prescription, characterized in that: The following steps are involved: Build a Chinese herbal medicine library and collect the chemical components of all Chinese herbal medicines in the library; Based on the chemical composition of all Chinese medicinal materials, the improved formula of compound Chinese medicine is obtained through computational analysis; The efficacy of the improved prescription was evaluated by calculating the correlation between the target set of the improved prescription and the disease gene set; The efficacy of the improved formula was verified through animal experiments; The improved formula of the compound Chinese medicine obtained by computational analysis based on the chemical components of all Chinese medicinal materials includes: Based on the chemical composition of all Chinese medicinal materials, a deep neural network (DNN) was used to build a Chinese medicinal material safety prediction model. The DNN model was used to predict the safety of adjuvants and adjuvants in compound Chinese medicines, and the safety prediction results of adjuvants and adjuvants were obtained. If the safety prediction of adjuvant and adjuvant drugs is extremely low, low or relatively low, the Chinese medicinal materials with extremely low, low or relatively low safety prediction are directly eliminated to obtain an improved prescription; If the safety prediction of adjuvant and adjuvant drugs is high, the improved formula is obtained after comprehensive analysis of the ADME properties of the Chinese medicinal materials and the target network correlation analysis; The ADME property analysis includes: Based on the chemical composition of Chinese medicinal materials, the names of the ingredients were standardized and duplicate data were eliminated to obtain the chemical composition of each Chinese medicinal material in the compound Chinese medicine formula; The chemical components of adjuvants and adjuvants were screened, eliminated, and ranked according to their ADME properties, with evaluation indicators including intestinal absorption, oral availability, and Lipinski rule; If the unique chemical components of adjuvant and adjuvant drugs in the compound Chinese medicine formula are the largest, then the Chinese medicine will be retained directly; The target network correlation analysis includes: The target of each Chinese herbal medicine in the compound Chinese medicine formula is obtained through text mining. After standardizing the target name and removing duplicate data, the corresponding target set is obtained. Based on the target set, the target network of the original compound Chinese medicine formula is constructed; all targets of the monarch and ministerial drugs in the compound Chinese medicine formula are retained, and the targets of the adjuvant or guiding drugs are randomly removed as the target set of the improved formula; After randomly removing the targets of adjuvant or adjuvant drugs, the centrality of the removed targets in the original prescription network and the importance in the key subnetwork were compared; After screening and sorting based on the ADME property analysis and the target network correlation analysis, the Chinese medicinal materials that need to be eliminated are finally obtained; after screening, the Chinese medicinal material ranked first is eliminated and used as an improved prescription, or the top two Chinese medicinal materials are eliminated at the same time and used as an improved prescription.

2. The method according to claim 1, characterized in that The bottom layer of the deep neural network DNN is the input layer, the middle layer is the hidden layer, and the last layer is the output layer. There are multiple hidden layers between the input layer and the output layer, and the layers are fully connected; the output of the input layer is represented by a vector χ, f(ω1χ+b1) is the output of the hidden layer, where ω1 is the weight, b1 is the bias, and the function f uses a sigmoid function, a relu function, or a tanh function; the output layer receives the output signal of the hidden layer and generates a prediction result, softmax(ω2χ1+b2) is the output of the output layer, where χ1 is the output of the hidden layer f(ω1χ+b1), and softmax converts the output of the neural network into a probability distribution That is, a Chinese herbal medicine X is judged to be of different safety categories The loss function used in training is the cross entropy loss function. All Chinese medicinal materials are divided into training set and test set. The model is trained and parameters are optimized through the training set. The accuracy of the model is evaluated through the test set. The evaluation indicators include F1score, accuracy and precision.

3. The method according to claim 1, characterized in that: The method of evaluating the efficacy of the improved prescription by calculating the correlation between the target set of the improved prescription and the disease gene set includes: Collect one or more disease genes related to compound Chinese medicine, and obtain the disease gene set after manual correction and removal of duplicate data; The random walk algorithm was used to calculate the correlation between the target set and the disease gene set of the original and improved formulas of compound Chinese medicine, and to evaluate the efficacy of the original and improved formulas on the disease, respectively.

Images