A hierarchical method, apparatus, equipment, and media for the mechanism of traditional Chinese medicine compound prescriptions based on the fusion graph network of the monarch-minister-adjuvant-guide theory.

By using a five-layer heterogeneous association network and the theory of principal, assistant, adjuvant, and guide drugs, combined with a single-cell gene regulatory network, the problem of target functional stratification in traditional network pharmacology was solved, realizing the functional stratification and biological validation of targets, and quantifying the importance and synergy of targets.

CN123120212APending Publication Date: 2026-11-17NAT INST FOR FOOD & DRUG CONTROL
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611598403.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-15
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional network pharmacology struggles to reveal the functional differences of targets in the overall therapeutic mechanism of traditional Chinese medicine (TCM) compound prescriptions, and topological index ranking alone cannot effectively quantify the importance of targets and cross-TCM synergy.

Method used

A five-layer heterogeneous association network was used to construct a traditional Chinese medicine compound. Based on the theory of monarch, minister, assistant and guide, the comprehensive node importance and node synergy of the target were calculated to screen the target segmentation threshold. The functional hierarchy of the target was verified by virtual knockout using a single-cell gene regulatory network.

Benefits of technology

This study achieved a clear stratification of target functions, quantified the core role of targets in the compound network and the support of multiple Chinese herbal medicines, reduced subjective bias in thresholds, and biologically verified the differences in target functions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123120212A_ABST
    Figure CN123120212A_ABST
Patent Text Reader

Abstract

This invention discloses a method, apparatus, equipment, and medium for hierarchical analysis of the mechanisms of traditional Chinese medicine (TCM) compound prescriptions based on a graph network fusion of the theory of monarch, minister, assistant, and guide herbs. It relates to the fields of TCM network pharmacology and bioinformatics. The method includes: constructing a five-layer heterogeneous association network and extracting the individual herb sub-networks corresponding to each component TCM herb; calculating the comprehensive node importance for target nodes and sequentially removing edges from the compound prescription network that contribute to the target target from each individual herb sub-network to calculate node synergy; traversing candidate partitioning thresholds, determining the target threshold based on the inter-group differences in the contribution of target therapeutic efficiency, and classifying targets into four categories (monarch, minister, assistant, and guide) according to comprehensive node importance and node synergy; performing virtual knockout of various targets based on a single-cell gene regulation network, verifying the hierarchical results based on the perturbation breadth and intensity, and analyzing the mechanisms of action of various targets through pathway enrichment analysis. This invention improves the interpretability of compound prescription mechanism analysis.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical fields of traditional Chinese medicine network pharmacology and bioinformatics, and in particular, relates to a method, apparatus, device and medium for stratifying the mechanism of a traditional Chinese medicine compound integrating a graph network based on the monarch-minister-assistant-guide theory. Background Art

[0002] Traditional Chinese medicine compounds are a commonly used therapeutic form in traditional Chinese medicine clinical practice, and their formulation follows the monarch-minister-assistant-guide compatibility theory, which realizes the overall regulation of the physiological processes of the organism by assigning different levels of functional positioning to multiple traditional Chinese medicines. Traditional Chinese medicine compounds usually have the characteristics of multi-component, multi-target and multi-pathway synergistic effects, and can involve complex biological processes such as metabolic regulation, immune modulation and intestinal flora remodeling.

[0003] Network pharmacology can construct interaction networks such as traditional Chinese medicine-compound-target, and use topological indicators to screen key targets. However, traditional topological indicators such as degree centrality mainly reflect the structural importance of nodes, and usually can only answer whether a certain target is important, making it difficult to further explain what function the target undertakes in the overall therapeutic mechanism of the compound. Simply prioritizing targets makes it difficult to reveal the functional differences between different targets in terms of main effects, synergistic connections, independent effects and background support.

[0004] The monarch-minister-assistant-guide theory includes a hierarchical description of the functional relationships of different components. Mapping this theory to the network target level can functionally classify targets from two dimensions: the contribution of the target to the overall compound network and the degree to which the target is jointly supported by multiple traditional Chinese medicines. Therefore, there is a need for a method that can quantify both target importance and cross-traditional Chinese medicine synergy on the basis of retaining the overall information of the compound network, and perform biological verification on the classification results. Summary of the Invention

[0005] The present invention aims to solve the problem that traditional network pharmacology mainly sorts targets by importance and lacks target functional stratification and verification, and provides a target functional hierarchy analysis solution guided by the monarch-minister-assistant-guide theory. This solution takes a five-layer heterogeneous association network as the data basis, takes the comprehensive node importance and node synergy of targets as two complementary dimensions, screens the target division threshold through the therapeutic efficiency contribution, and combines the single-cell gene regulatory network virtual knockout to verify the stratification results.

[0006] In a first aspect, the present invention provides a method for stratifying the mechanism of a traditional Chinese medicine compound integrating a graph network based on the monarch-minister-assistant-guide theory, the method comprising: Obtaining the constituent traditional Chinese medicines of the traditional Chinese medicine compound, as well as compounds, targets, intestinal flora and flora metabolites associated with the constituent traditional Chinese medicines, constructing a five-layer heterogeneous association network according to the association relationships between different types of nodes, and extracting a single-traditional Chinese medicine sub-network corresponding to each constituent traditional Chinese medicine from the five-layer heterogeneous association network; For each target node in the five-layer heterogeneous association network, the degree centrality, betweenness centrality, proximity centrality, and PageRank value are calculated. The calculation results are sorted and rank sum is performed to obtain the comprehensive node importance of each target node. For each target node, the edges contributed by each single herb sub-network to the corresponding target node are removed sequentially in the five-layer heterogeneous association network. The residual degree centrality of the corresponding target node after each removal is calculated, and the node collaboration of the corresponding target node is obtained by combining the original degree centrality before removal. Traverse the candidate partitioning thresholds, and under each candidate partitioning threshold, classify each target node into four categories based on the importance of the comprehensive node and the synergy of the nodes. Calculate the treatment efficiency contribution of each target node, and determine the target partitioning threshold based on the difference in treatment efficiency contribution between different categories of targets. Based on the target segmentation threshold and the monarch-minister-assistant-envoy theory, each target node is divided into monarch target, minister target, assistant target and envoy target, to obtain the target functional hierarchy system. Gene regulatory networks were constructed based on single-cell transcriptome data of the target disease. Virtual knockout was performed on the principal, assistant, adjuvant, and agonist targets, respectively. The functional hierarchy of the targets was verified based on the perturbation breadth and intensity after virtual knockout.

[0007] Furthermore, the five-layer heterogeneous association network includes a traditional Chinese medicine layer, a compound layer, a target layer, a gut microbiota layer, and a microbiota metabolite layer. The relationships between traditional Chinese medicine and compounds, and between compounds and targets, are derived from the HERB database. The relationships between targets and gut microbiota, targets and microbiota metabolites, and microbiota metabolites and gut microbiota are derived from the gutMGene database. The relationships between targets are derived from the STRING database. The remaining cross-layer relationships are derived from literature searches.

[0008] Furthermore, the overall node importance INI of target node j is calculated according to the following formula: in, , , and These represent the degree centrality ranking, betweenness centrality ranking, proximity centrality ranking, and PageRank ranking of target node j, respectively. The smaller the INI value, the higher the overall node importance of the corresponding target node.

[0009] Furthermore, the node cooperation NS of target node j is calculated according to the following formula: in, The quantity of the constituent Chinese medicines, For the i-th single herb network, To remove the original degree centrality of the former target node j in the five-layer heterogeneous association network, To remove the edge contribution of the i-th single-herb sub-network to target node j from the five-layer heterogeneous association network, the residual degree centrality of target node j is determined; when target node j does not exist in the i-th single-herb sub-network, Take 0.

[0010] Further, determining the target segmentation threshold includes: traversing candidate percentage thresholds in a step size of 1% within the range of 5% to 25%; classifying targets into principal, assistant, adjuvant, and adjuvant categories under each candidate percentage threshold, and calculating the therapeutic efficiency contribution after single target side knockout; comparing the average therapeutic efficiency contribution difference between the principal target and the other three types of targets and performing statistical tests; and determining the candidate percentage threshold that meets the statistical significance requirement between groups and can increase the functional difference between categories as the target segmentation threshold.

[0011] Furthermore, according to the target classification threshold, targets with high overall node importance and high node synergy are classified as principal targets, targets with low overall node importance and high node synergy are classified as subordinate targets, targets with high overall node importance and low node synergy are classified as assistant targets, and targets with low overall node importance and low node synergy are classified as enable targets.

[0012] Furthermore, the virtual knockout process uses scTenifoldKnk to process the single-cell gene expression counting matrix of the target cell population, constructs a baseline gene regulation network, and generates virtual knockout networks for each type of target. The number of significantly perturbed genes and the average fold change of significantly perturbed genes are obtained through manifold alignment and differential regulation analysis, which are used as the perturbation breadth and the perturbation intensity, respectively. The method also includes using the clusterProfiler package to perform pathway enrichment analysis on all targets, as well as the principal, assistant, adjuvant, and facilitator targets.

[0013] Secondly, this invention provides a hierarchical device for the mechanism of traditional Chinese medicine compound prescriptions based on a fusion graph network of the theory of principal, assistant, adjuvant, and guide herbs. The device is composed of program modules corresponding to the steps of the aforementioned method. The device is a functional module architecture for implementing the data processing flow and is not limited to dedicated physical hardware.

[0014] Thirdly, the present invention provides an electronic device including a processor and a memory, wherein the processor implements the above-described method when executing a computer program in the memory or when the computer executes instructions.

[0015] Fourthly, the present invention provides a computer-readable storage medium for storing a corresponding computer program or computer execution instructions.

[0016] This invention has at least the following beneficial effects: First, it extends the target topological sorting of traditional network pharmacology to a target hierarchy with clear functional meaning; second, it characterizes the core degree of a target in the complete compound network and the degree to which it is supported by multiple Chinese medicines through two complementary dimensions: node importance and node synergy; third, it determines the target partitioning threshold by traversing candidate thresholds and comparing the differences in the contribution of treatment efficiency, reducing the bias caused by subjective threshold setting; fourth, it uses the virtual knockout results of single-cell gene regulation networks to verify the differences in biological functions of different types of targets from two dimensions: perturbation breadth and perturbation intensity. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 A schematic diagram of the five-layer heterogeneous correlation network and the network of eight single herbs in the traditional Chinese medicine formula for clearing the intestines and warming the body. Figure 3 A comparison chart showing the ranking of the top 20 targets in terms of overall importance in the compound drug network and the internal ranking of each individual herb in the sub-network; Figure 4 This is a schematic diagram illustrating the principle of node collaborative computing. Figure 5 A comparison chart of the top 10 targets in terms of overall node importance and the ranking of node synergy; Figure 6 Curves showing the changes in the average therapeutic efficiency contribution of four target classes under different candidate thresholds; Figure 7 Curves showing the difference in the contribution of the target to the average treatment efficiency of other categories under different candidate thresholds; Figure 8 A statistical chart showing the contribution of four target categories to treatment efficiency under different thresholds. Figure 9 A four-category partitioning diagram for all targets based on a comprehensive assessment of node importance and node synergy; Figure 10 A diagram showing cell clustering and differentially expressed gene analysis for single-cell data; Figure 11 Heatmap and fold change plot of the perturbation gene after virtual knockout; Figure 12 A comparison chart of the perturbation extent of different target categories; Figure 13 A comparison chart of perturbation intensity for different types of targets; Figure 14 This is a pathway enrichment analysis diagram for all targets and four types of targets. Detailed Implementation

[0018] The following example illustrates the invention using the Qingchang Wenzhong formula for treating ulcerative colitis, but this example does not constitute a limitation on the scope of protection. Those skilled in the art can implement this invention using the same target importance-synergy dual-dimensional analysis framework for other traditional Chinese medicine formulas and target diseases.

[0019] In this embodiment, "target" refers to the protein or gene product of the compound that acts in the body; "single herb sub-network" refers to a subgraph that takes a certain component of traditional Chinese medicine as its source and includes the traditional Chinese medicine and its associated compounds, target, intestinal flora, flora metabolites and corresponding edge relationships; "target treatment efficiency contribution" refers to the change in network treatment efficiency relative to before removal after removing all edges of the target in the compound network.

[0020] This invention provides a method for hierarchical analysis of the mechanisms of traditional Chinese medicine compound prescriptions based on a graph network fusion of the theories of monarch, minister, assistant, and guide. Figure 1 As shown, the method includes the following steps S10-S70.

[0021] Step S10: Network construction and single herb network extraction.

[0022] Data on the constituent Chinese herbal medicines, compounds, targets, gut microbiota, and microbiota metabolites of traditional Chinese medicine (TCM) formulas were obtained. A five-layer heterogeneous association network was constructed, and individual herb sub-networks corresponding to each constituent Chinese herbal medicine were extracted based on the TCM source of nodes and edges. Here, the five-layer heterogeneous association network refers to a heterogeneous graph that includes five types of nodes and multiple types of cross-layer or intra-layer relationships, rather than a complete graph in the sense of graph theory where any two nodes are directly connected.

[0023] This embodiment uses a traditional Chinese medicine formula for clearing the intestines and warming the body as an example. Its publicly disclosed components include dried ginger, licorice root, Panax notoginseng, Sophora flavescens, Coptis chinensis, Aucklandia lappa, Sanguisorba officinalis, and Indigo naturalis. Compound data were obtained from the HERB database; the intersection of compound targets with ulcerative colitis disease targets obtained from the GeneCards, TTD, and OMIM databases was taken; gut microbiota and microbiota metabolites related to the targets were obtained from the gutMGene database; interactions between targets were obtained from the STRING database; other cross-layer relationships were obtained through PubMed and CNKI literature searches. To reduce network noise, isolated nodes without cross-layer connections were removed.

[0024] The processed nodes and edges were imported into the Neo4j graph database. Each type of biological entity was defined as a corresponding node, and known associations between entities were defined as edges. A compound network was generated using Cypher queries, and eight single-herb sub-networks were extracted based on the origin of the nodes and edges. Overlapping nodes and edges are allowed between the single-herb sub-networks; this overlap relationship forms the structural basis for subsequent calculations of target synergy.

[0025] Step S20: Calculate the importance of target nodes.

[0026] For each target node in the complex network, the degree centrality, betweenness centrality, proximity centrality, and PageRank values ​​are calculated and sorted from largest to smallest. The overall node importance INI of target node j is calculated according to the following formula: in, , , and These represent the ranking of target node j in the corresponding indicators. INI integrates local connectivity, path control capability, global reachability, and overall propagation influence using a rank-sum method. The smaller the INI value and the higher the ranking, the higher the overall node importance of the target.

[0027] Step S30: Calculation of target node collaboration.

[0028] For target node j, remove all edges from the complete compound herb sub-network hi that contribute to the target node, while retaining the edges related to the target node with other single herb sub-networks. Calculate the residual centrality of target node j after each removal. The node collaboration NS of target node j is calculated according to the following formula: Where m represents the quantity of the constituent Chinese medicinal herbs, in this embodiment... It equals 8; The degree centrality of target node j in the original compound network; when target node j does not exist in the single herb sub-network. China Times, Take 0. For example... Figure 4 As shown, NS is calculated by removing the edges contributed by each individual herb in the sub-network. The higher the NS, the more the target is still supported by the action pathways of other herbs after removing the relevant edges of any herb, indicating a stronger cross-herb mechanism overlap synergy.

[0029] Step S40, target threshold filtering.

[0030] The target INI and NS are mapped to the functional logic of the monarch-minister-assistant-envoy theory. To reduce the bias caused by manually specified thresholds, candidate percentage thresholds are traversed in steps of 1% within the range of 5% to 25%. Under each candidate threshold, targets with INI rankings within the threshold range are defined as high-importance targets, and targets with NS rankings within the threshold range are defined as high-synergy targets, thus forming four types of targets.

[0031] Therapeutic efficiency (TE) is used to measure the structural efficiency of a complex network in transmitting biological signals from therapeutic layer nodes to disease layer targets. Therapeutic layer nodes include traditional Chinese medicine, compounds, microbial metabolites, and gut microbiota, while disease layer nodes are the targets. TE is calculated using the following formula: in, and Let represent the number of nodes in the treatment layer and the number of nodes in the disease layer, respectively, and let dij be the shortest path distance from treatment layer node i to disease layer target point j. Then, remove all edges from the target point and recalculate the treatment efficiency, and the target point's contribution to the treatment efficiency. Calculate according to the following formula: For each candidate threshold, calculate the average of the four target classes. And compare the average values ​​between the monarch target, minister target, adjuvant target, and envoy target. Differences. One-way ANOVA and Tukey's test were used to evaluate differences between groups. Among the candidate thresholds that met the statistical significance requirements, the target threshold was determined by combining the functional differences between categories and the number of target points in each category.

[0032] Step S50: Construct the target functional hierarchy system TFHS based on INI and NS.

[0033] A target functional hierarchy system (TFHS) is constructed based on target thresholds. The principal target has high importance and high synergy, serving as the core driving target in the compound drug network; the assistant target has lower importance and high synergy, acting as a synergistic hub connecting different pathways of action of traditional Chinese medicine; the adjuvant target has high importance and lower synergy, driven more by specific single herbs and playing a relatively independent role; thus, targets with lower importance and synergy constitute the background connections supporting network operation.

[0034] Step S60: Virtual knockout verification of single-cell gene regulatory network.

[0035] Raw single-cell transcriptome counts were obtained from patients with the target disease and healthy controls. Cell quality control, standardization, and cell population segmentation were performed. Target cell populations related to the pathological process of the target disease were selected, and differentially expressed genes between the patient group and the healthy control group were screened. Using the gene expression count matrix of the target cell population as input, a baseline gene regulatory network was constructed using scTenifoldKnk. Virtual knockouts were performed on the principal, assistant, adjuvant, and agonist targets, and the gene perturbation results after knockout were obtained through tensor dimensionality reduction, virtual knockout network generation, manifold alignment, and differential regulation analysis.

[0036] This embodiment uses the default parameters of the scTenifoldKnk software, including a mitochondrial reads proportion threshold of 0.1, a minimum library size of 1000, a network count of 10, a cell sampling count per network of 500, a principal component count of 3, an edge relation quantile threshold of 0.9, a tensor rank of 3, a maximum number of iterations of 1000, and a relative error threshold of 1×10⁻⁶. -5 And manifold dimension 2. For data that does not meet the default quality control conditions, the corresponding parameters can be adjusted according to the data scale and sequencing depth without changing the overall process of this invention.

[0037] The number of perturbation genes that met the software's significance criteria after virtual knockout of each target was counted and used as the perturbation breadth (Sig_count); the average fold change (Mean_FC) of significantly perturbation genes was calculated as the perturbation intensity. The perturbation breadth and intensity of the four target classes were compared to verify the different functional roles defined by TFHS.

[0038] Step S70, pathway enrichment analysis.

[0039] The KEGG pathway enrichment function of the clusterProfiler package in R was used to perform pathway enrichment analysis on all targets of the compound, as well as the principal, assistant, adjuvant, and guide targets. Multiple validation corrections were applied to the enrichment results, and significant pathways related to the target disease were screened to compare the functional differences of different target categories in core inflammation, immune regulation, cell survival, hypoxia adaptation, and other biological processes.

[0040] Example results: A five-layer heterogeneous relational network for the treatment of ulcerative colitis using the Qingchang Wenzhong formula was constructed in the Neo4j database. This network contains 1076 nodes and 8593 edges, and eight single-herb subnetworks were extracted from it, such as... Figure 2 As shown in Table 1, the network construction data collection results are presented. The nodes and edges of the networks for dried ginger, licorice, Panax notoginseng, Sophora flavescens, Coptis chinensis, Aucklandia lappa, Sanguisorba officinalis, and Indigo naturalis exhibit varying degrees of overlap, providing a structural basis for cross-traditional Chinese medicine collaborative computation at the target level.

[0041] Table 1 Data Collection Results

[0042] Calculate the target INI in the complete compound network and the sub-networks of each single herb. For example... Figure 3 As shown, the top-ranked targets in the compound drug network are not simply the sum of contributions from individual herb subnetworks. For example, PTGS2 ranks first in the compound drug network, but its ranking is relatively low in most individual herb subnetworks, indicating that its high topological position comes from the interaction between multiple Chinese herbal medicines.

[0043] Further calculate the target NS. For example... Figure 5 As shown, comparing the NS rankings of the top 10 targets in the INI ranking reveals that target importance and synergy are different attributes. For example, TNF ranks 4th in the INI and 65th in the NS ranking, while IL6 ranks 5th in the INI and 87th in the NS ranking. Therefore, it is necessary to combine both dimensions for functional stratification.

[0044] After iterating through the candidate thresholds from 5% to 25%, as follows: Figure 6 As shown, the average TECtarget of the target is the highest overall under each candidate threshold. Figure 7 As shown, the functional differences between the 5% and 8% thresholds are significant, but the number of target points (monarch, minister, and assistant) is relatively small at the 5% threshold, which is not conducive to subsequent statistical analysis. Therefore, 8% is determined as the target threshold in this embodiment. At the 8% threshold, as... Figure 9 As shown, the 270 target points are divided into 10 monarch targets, 11 minister targets, 11 adjuvant targets, and 238 messenger targets, as follows: Figure 8 As shown, the differences in TECtarget among the four target types reached a statistically significant level.

[0045] Virtual knockout validation was performed using the ulcerative colitis single-cell dataset GSE214695. This dataset includes 6 ulcerative colitis samples and 6 healthy controls. Figure 10 As shown, after quality control and cell segmentation, intestinal epithelial cells were selected, and 367 differentially expressed genes were obtained. Due to computational resource limitations, 60 genes were randomly selected from 238 target sites for virtual knockout, and all principal, assistant, and adjuvant target sites were included in the analysis.

[0046] The virtual knockout results show that, Figure 11 As shown, virtual knockout produces different perturbation genes and fold changes in different target categories. Regarding the breadth of perturbation, as... Figure 12 As shown, there was no significant difference between the monarch target and the minister target, both of which were higher than the assistant target and the facilitator target; regarding the perturbation intensity, as Figure 13 As shown, there is no significant difference between the monarch target and the assistant target; both are higher than the minister target and the facilitator target. The perturbation intensity of the minister target is higher than that of the facilitator target. This result corresponds to the theoretical expectation that the monarch target has a wide and strong perturbation, the minister target has a wide but relatively mild perturbation, the assistant target has a narrow range but a high intensity perturbation, and the facilitator target has a narrow and weak perturbation.

[0047] Pathway enrichment results show that, for example Figure 14As shown, all targets of the compound are mainly enriched in the IL-17 signaling pathway, inflammatory bowel disease pathway, Toll-like receptor pathway, TNF signaling pathway, and HIF-1 signaling pathway. The principal target mainly involves core inflammatory signals and epithelial cell fate regulation, the assistant target mainly involves angiogenesis, hypoxia adaptation, and cell survival, the adjuvant target mainly involves ulcerative colitis-specific immune disorders, and the final target mainly involves a broad range of general inflammatory signals. The pathways of various targets are both distinct and partially overlapping, supporting the functional stratification of TFHS targets at the mechanistic level.

[0048] The above embodiments illustrate that this invention, through target-based node importance, node synergy, target threshold screening, and single-cell virtual knockout verification, can form a target functional hierarchy with the implications of the traditional Chinese medicine theory of monarch, minister, assistant, and guide. The five-layer heterogeneous association network provides the data foundation for the analysis. The key innovations of this invention lie in the dual-dimensional evaluation of targets, data-driven threshold determination, and biological verification of the hierarchical results.

Claims

1. A hierarchical method for the mechanism of traditional Chinese medicine compound prescriptions based on a graph network fusion of the theory of monarch, minister, assistant, and guide, characterized in that, include: The constituent Chinese medicines of a traditional Chinese medicine compound are obtained, along with the compounds, targets, gut microbiota, and microbiota metabolites associated with the constituent Chinese medicines. A five-layer heterogeneous association network is constructed based on the association relationships between different types of nodes, and the single herb sub-networks corresponding to each constituent Chinese medicine are extracted from the five-layer heterogeneous association network. For each target node in the five-layer heterogeneous association network, the degree centrality, betweenness centrality, proximity centrality, and PageRank value are calculated. The calculation results are sorted and rank sum is performed to obtain the comprehensive node importance of each target node. For each target node, the edges contributed by each single herb sub-network to the corresponding target node are removed sequentially in the five-layer heterogeneous association network. The residual degree centrality of the corresponding target node after each removal is calculated, and the node collaboration of the corresponding target node is obtained by combining the original degree centrality before removal. Traverse the candidate partitioning thresholds, and under each candidate partitioning threshold, classify each target node into four categories based on the importance of the comprehensive node and the synergy of the nodes. Calculate the treatment efficiency contribution of each target node, and determine the target partitioning threshold based on the difference in treatment efficiency contribution between different categories of targets. Based on the target segmentation threshold and the monarch-minister-assistant-envoy theory, each target node is divided into monarch target, minister target, assistant target and envoy target, to obtain the target functional hierarchy system. Gene regulatory networks were constructed based on single-cell transcriptome data of the target disease. Virtual knockout was performed on the principal, assistant, adjuvant, and agonist targets, respectively. The functional hierarchy of the targets was verified based on the perturbation breadth and intensity after virtual knockout.

2. The method according to claim 1, characterized in that, The five-layer heterogeneous association network includes a traditional Chinese medicine layer, a compound layer, a target layer, a gut microbiota layer, and a microbiota metabolite layer. The relationships between traditional Chinese medicine and compounds, and between compounds and targets, are derived from the HERB database. The relationships between targets and gut microbiota, targets and microbiota metabolites, and microbiota metabolites and gut microbiota are derived from the gutMGene database. The relationships between targets are derived from the STRING database. The remaining cross-layer relationships are derived from literature searches.

3. The method according to claim 1, characterized in that, The overall node importance INI of target node j is calculated according to the following formula: in, , , and These represent the degree centrality ranking, betweenness centrality ranking, proximity centrality ranking, and PageRank ranking of target node j, respectively. The smaller the INI value, the higher the overall node importance of the corresponding target node.

4. The method according to claim 1, characterized in that, The node collaboration NS of target node j is calculated according to the following formula: in, The quantity of the constituent Chinese medicines, For the i-th single herb network, To remove the original degree centrality of the former target node j in the five-layer heterogeneous association network, To remove the edge contribution of the i-th single-herb sub-network to target node j from the five-layer heterogeneous association network, the residual degree centrality of target node j is determined; when target node j does not exist in the i-th single-herb sub-network, Take 0.

5. The method according to claim 1, characterized in that, Determining the target segmentation threshold includes: traversing candidate percentage thresholds in a step size of 1% within the range of 5% to 25%; classifying targets into principal, assistant, adjuvant, and adjuvant categories under each candidate percentage threshold, and calculating the therapeutic efficiency contribution after single target side knockout; comparing the average therapeutic efficiency contribution difference between the principal target and the other three target categories and performing statistical tests; and determining the candidate percentage threshold that meets the statistical significance requirement between groups and can increase the functional difference between categories as the target segmentation threshold.

6. The method according to claim 1, characterized in that, According to the target classification threshold, targets with high overall node importance and high node synergy are classified as "King" targets, targets with low overall node importance and high node synergy are classified as "Minister" targets, targets with high overall node importance and low node synergy are classified as "Assistant" targets, and targets with low overall node importance and low node synergy are classified as "Make" targets.

7. The method according to claim 1, characterized in that, The virtual knockout method uses scTenifoldKnk to process the single-cell gene expression counting matrix of the target cell population, constructs a baseline gene regulation network, and generates virtual knockout networks for each type of target. The number of significantly perturbed genes and the average fold change of significantly perturbed genes are obtained through manifold alignment and differential regulation analysis, which are used as the perturbation breadth and the perturbation intensity, respectively. The method also includes using the clusterProfiler package to perform pathway enrichment analysis on all targets as well as the principal, assistant, adjuvant, and facilitator targets.

8. A hierarchical device for the mechanism of traditional Chinese medicine compound prescriptions based on a fusion graph network of the theory of monarch, minister, assistant, and guide, characterized in that, include: The module includes a network construction module, an importance calculation module, a collaborative calculation module, a threshold screening module, a target hierarchical module, and a virtual knockout verification module. The network construction module is used to construct a five-layer heterogeneous association network and extract the single-herb sub-networks corresponding to each component Chinese medicine; the importance calculation module is used to calculate the comprehensive node importance of each target node; the synergy calculation module is used to perform edge knockout in the five-layer heterogeneous association network according to the single-herb sub-networks and calculate the node synergy of each target node; the threshold screening module is used to traverse candidate partitioning thresholds and determine the target partitioning threshold based on the difference in the contribution of target treatment efficiency; the target hierarchical module is used to construct a target functional hierarchy system based on the comprehensive node importance, the node synergy, and the target partitioning threshold; the virtual knockout verification module is used to verify the target functional hierarchy system based on a single-cell gene regulation network.

9. An electronic device, characterized in that, It includes a processor and a memory communicatively connected to the processor; the memory stores a computer program or computer execution instructions, and when the processor executes the computer program or computer execution instructions, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program or computer-executable instructions thereon, characterized in that, When the computer program or computer execution instructions are executed by the processor, the method of any one of claims 1 to 7 is implemented.