Search

CN-121983249-A - Pharmacological analysis system for traditional Chinese medicine prescription

CN121983249ACN 121983249 ACN121983249 ACN 121983249ACN-121983249-A

Abstract

The invention relates to the technical field of information retrieval, and provides a pharmacological analysis system for a traditional Chinese medicine prescription, which comprises a plurality of groups of chemical data sets of a plurality of groups of chemical feature fusion subsystems, wherein the groups of chemical feature vectors are obtained after feature extraction and dimension reduction operation, the groups of chemical feature vectors capture the comprehensive influence of traditional Chinese medicine prescriptions on genome, proteome and metabolome levels, the component-target dynamic interaction network of a component synergistic network construction subsystem is subjected to clustering and community detection to obtain a component synergistic interaction sub-network, the prescription-disease association map of a disease association path discovery subsystem is subjected to path mining to obtain a potential disease association path set, a novel prescription-disease association path is obtained, and the traditional Chinese medicine prescription optimization candidate scheme of a prescription optimization decision generation subsystem is subjected to multi-objective optimization to obtain a traditional Chinese medicine prescription optimization decision set. The invention realizes the full-chain association analysis of chemical components, action targets, biological pathways and treatment diseases of the traditional Chinese medicine prescription.

Inventors

  • BAI WEIMIN

Assignees

  • 安顿健康科技有限公司

Dates

Publication Date
20260505
Application Date
20251211

Claims (10)

  1. 1. A pharmacological analysis system for a pharmaceutical formulation, comprising: The component collaborative network construction subsystem is used for carrying out network analysis processing on a plurality of groups of chemical feature vectors containing structured data and unstructured data, calculating the association strength between chemical components and constructing a component-target dynamic interaction network; the system comprises a disease association path discovery subsystem, a prescription-disease association path acquisition subsystem and a prescription-disease association path analysis subsystem, wherein the disease association path discovery subsystem is used for integrating a composition synergistic interaction sub-network with a disease knowledge base through the treatment of a knowledge graph construction module to form a prescription-disease association graph; The prescription optimization decision generation subsystem is used for obtaining a traditional Chinese medicine prescription optimization candidate scheme by simulating traditional Chinese medicine prescription compatibility adjustment and effect prediction through a novel prescription-disease association path, machine learning processing and prescription optimization function, and obtaining a traditional Chinese medicine prescription optimization decision set through multi-objective optimization of the traditional Chinese medicine prescription optimization candidate scheme.
  2. 2. The pharmacological analysis system for pharmaceutical formulation according to claim 1, wherein the component collaborative network construction subsystem comprises: The dynamic interaction relation inference component is used for obtaining a conditional association probability distribution between the chemical components and the target protein through conditional dependence analysis based on multiple groups of chemical contexts on multiple groups of chemical feature vectors, and constructing a component-target initial association skeleton through significance filtering based on statistical potential energy on the conditional association probability distribution; The network topology structure generation component is used for carrying out connection strength assignment by introducing time sequences or dose response modes contained in a plurality of groups of the mathematical feature vectors to form a weighted component-target dynamic interaction network; The functional community discovery component is used for obtaining potential functional community division in the network through functional module boundary detection based on network flow betweenness centrality by the weighted component-target dynamic interaction network obtained by the network topology structure generation component, and the potential functional communities form a component synergistic sub-network through optimized screening based on module cohesiveness and module separation.
  3. 3. The pharmacological analysis system for pharmaceutical formulation according to claim 1, wherein the disease associated pathway discovery subsystem comprises: The heterogeneous knowledge network cooperative reinforcement component is used for forming a preliminarily integrated heterogeneous map by processing the component cooperative sub-network through the knowledge map construction module, aligning with the entity of the disease knowledge base and mapping the relationship; the heterogeneous atlas models and complements potential higher-order relations between component-target cooperative groups and target-disease association by introducing relation complementation and cooperative embedding learning based on a graph neural network, and a cooperative enhancement knowledge network is generated; The causal mechanism path reasoning component is used for exploring the mode recognition capability of the graph annotation force network by combining rules of inductive logic programming through the processing of a symbol and numerical value based hybrid reasoning engine and traversing and generating a candidate action path sequence for connecting a specific prescription cooperative group and a disease node in the knowledge network; the multi-dimensional evidence aggregation scoring and screening component is used for quantitatively evaluating and weighting and fusing four dimensions of path compactness, enrichment significance of path related genes on related disease paths, disturbance correlation of histology data on path key nodes and document co-occurrence support degree through processing of a multi-dimensional evidence aggregation scoring mechanism of a prescription-disease candidate mechanism path, calculating comprehensive confidence score of each path, sorting all paths according to the comprehensive confidence scores, and screening based on dynamic thresholds to output a novel prescription-disease association path with high confidence.
  4. 4. The pharmacological analysis system for a pharmaceutical formulation of claim 3, wherein the causal mechanism pathway inference component comprises: The causal hypothesis generation sub-component is used for generating a series of initial meta-path example sets for connecting the prescription collaborative group and the diseases through matching and instantiation processing based on a high-order relation template by the collaborative enhancement knowledge network, wherein the initial meta-path example sets are filled and checked by entities based on network embedded similarity to form initial causal path hypothesis sets from fine granularity to specific entities; The system comprises a path optimization sub-component, a symbol reasoner, a neural network reasoner, a collaborative optimization algorithm, a logic enhancement and confidence enhancement optimization causal path set, a logic enhancement and semantic enhancement optimization causal path set and a logic enhancement and confidence enhancement optimization causal path set, wherein the path optimization sub-component is used for processing an initial causal path hypothesis set through a neural symbol joint reasoning engine; The biological pruning sub-assembly is used for optimizing a causal path set to be subjected to pruning processing based on dynamic biological constraint, calculating a constraint conflict score according to the degree of each optimizing causal path violating the dynamic constraint, sorting and threshold screening all paths according to the constraint conflict score, and finally outputting a prescription-disease candidate mechanism path.
  5. 5. The pharmacological analysis system for chinese medicine as recited in claim 4, wherein the biological pruning subassembly comprises: The time-space dependency graph construction module is used for performing relationship standardization processing on organelle co-localization data and biological process time sequence data in dynamic biological constraint to form a standardized space co-localization relationship set and a time sequence relationship set; In the space verification channel, the interaction relation of adjacent entity pairs in the causal path is optimized, and a series of space consistency verification results are generated by comparing the space co-localization probabilities of the adjacent entity pairs with the space co-localization probabilities of corresponding entities in the space-time dependency graph; the system comprises a conflict severity quantification and integration module, a time-space constraint conflict instance set evaluation module and a constraint conflict score generation module, wherein each space conflict instance carries out severity assignment according to a co-location probability value of an entity pair, and all assigned conflict instances are integrated to generate a constraint conflict score representing that the whole path violates biological rationality through calculation based on a nonlinear aggregation function.
  6. 6. The pharmacological analysis system according to claim 5, wherein the conflict severity quantification and integration module comprises: The conflict-dependent network construction submodule is used for carrying out analysis on the time-space constraint conflict instance set which is assigned, and identifying the initiated, aggravated or concurrent dependency relationship existing between the instances based on the causal and conditional dependency relationship among the conflicts; The cascade failure effect simulation sub-module is used for iteratively calculating the propagation process and energy accumulation of the collision influence along the network topology structure according to the dependency intensity and type represented by the edge from the initial collision node in the network by the collision dependency network simulation, obtaining a cascade influence value overlapped by the upstream collision propagation of the network after multiple rounds of simulation; and the comprehensive conflict score generation sub-module is used for carrying out weighted fusion on the initial severity assignment of each conflict node and the cascade influence value of each conflict node to obtain the comprehensive conflict strength of the nodes, and generating the comprehensive conflict score as the constraint conflict score.
  7. 7. The pharmacological analysis system for chinese medicine as recited in claim 6, wherein said composite conflict score generation submodule comprises: The system comprises a conflict strength grade dividing unit, a grade identifier, a grade classification unit and a grade classification unit, wherein the conflict strength grade dividing unit is used for dividing the comprehensive conflict strength of each node in a conflict state update set into a plurality of predefined conflict strength grades through grade judgment processing based on a dynamic threshold value; the cross-grade strength transfer unit is used for a conflict node set with a grade structure, and is processed by a cross-grade strength transfer function, wherein the sum of the original strengths of nodes in each grade and the received transfer strength from a higher grade are synthesized by vector addition to generate a group of grade strength vectors after transfer correction; The saturated nonlinear aggregation unit is used for calculating the modular length of the grade intensity vector as an aggregation input value, substituting the aggregation input value into a monotonically increasing function with an upper asymptote for mapping, and generating a final constraint conflict fraction by standardized scaling of the mapped output value.
  8. 8. The pharmacological analysis system for pharmaceutical formulations according to claim 7, wherein the saturated nonlinear polymerization unit comprises: the scale reference construction subunit is used for constructing a theoretical reference upper limit and a statistical reference distribution required by the standardization process through analysis based on a theoretical maximum output value and an empirical distribution percentile; the dynamic scale alignment subunit is used for monotonously increasing the output value of the function, the theoretical reference upper limit and the statistical reference distribution generated by the scale reference construction unit, and performing scale conversion processing based on nonlinear interpolation; the score interval calibration subunit is used for generating an intermediate scale value by the dynamic scale alignment unit, carrying out linear affine transformation processing based on a preset target interval, accurately adjusting the value range of the intermediate scale value to be within the preset target value interval through translation and scaling operation, and generating a final adjusted value which is a standardized constraint conflict score with comparative significance.
  9. 9. The pharmacological analysis system for pharmaceutical formulation according to claim 1, wherein the formulation optimization decision-making subsystem comprises: The candidate disturbance generation component is used for identifying a target point group and a biological process which have core regulation and control effects on a target disease mechanism through analysis processing based on path key nodes on a novel prescription-disease association path, comparing an identification result with a component synergistic effect sub-network of an original prescription, and generating a group of increase-decrease-adjustment initial operation instruction set aiming at prescription components by calculating the network structure difference and node function coincidence degree; The system comprises a primary optimization scheme set, a treatment effect simulation and evaluation component, a chemical stability channel, a combination coordination channel, a three-channel output and analysis module, wherein the primary optimization scheme set passes through three core evaluation channels of the multi-mode depth treatment effect simulator, the three core evaluation channels are used for simulating downstream biological effect changes after acting on a disease association path in a pharmacological effect channel through a graph neural network model trained based on biomedical knowledge graph and molecular docking prediction data and outputting predicted pesticide effect strength; the pareto front navigation component is used for carrying all schemes of a quantitative evaluation report, entering a multi-target optimization space, wherein defined core optimization targets in the space comprise maximization of predicted efficacy strength, minimization of potential stability risk and minimization of deviation of compatibility characteristics with a primary agent core, the optimization process excludes illegal schemes according to rigid constraint of a clinical safety threshold, searches schemes which cannot be continuously improved on any one target and cannot damage other targets in the remaining schemes to form a pareto optimal front, and selects a prescription optimization decision set from the optimal front through front search based on decision maker preference.
  10. 10. The pharmacological analysis system for pharmaceutical formulation according to claim 1, further comprising a multi-group feature fusion subsystem for providing multi-source data comprising structured data and unstructured data, wherein the multi-group feature fusion subsystem is configured to obtain a multi-group feature data set through multi-group integrated analysis processing, and wherein the multi-group feature data set is configured to obtain a fused multi-group feature vector through feature extraction and dimension reduction operations.

Description

Pharmacological analysis system for traditional Chinese medicine prescription Technical Field The invention relates to the technical field of information retrieval, in particular to a pharmacological analysis system for a traditional Chinese medicine prescription. Background The traditional Chinese medicine prescription is used as a core carrier of the theory and practice of the traditional Chinese medicine, has the characteristics of multiple components, multiple targets and integral regulation, but faces a plurality of technical challenges in the process of modern pharmacological research, firstly, the chemical components of the traditional Chinese medicine prescription are complex, the effective components are unclear, the drug effect substance basis is difficult to be clear, secondly, the interaction between the traditional Chinese medicine prescription and the organism relates to a plurality of biological layers, the explanation of the action mechanism is insufficient from molecules, cells, tissues to organs, and furthermore, the traditional Chinese medicine prescription research is mostly dependent on the personal experience of doctors, and a standardized and quantified analysis method is lacked. Along with the continuous development and progress of science and technology, the existing digital traditional Chinese medicine platform has a certain progress in data integration and simple analysis, but has obvious defects in aspects of multi-source data fusion, intelligent analysis model construction, action mechanism system analysis and the like. for example, existing platforms often have difficulty in achieving full-scale associative analysis from prescription chemistry to disease treatment, and lack deep mining and visualization capabilities for prescription-composition-target-pathway-disease multi-level networks. The prior art CN120561281A relates to an intelligent matching system for traditional Chinese medicine formula, which comprises a data acquisition module, a data processing module, a model training module and a prescription matching module, wherein the data acquisition module acquires medicine and formula information from medical documents, medical databases, experimental research results and clinical cases through a data interface and a crawler technology and stores the medicine and formula information into a unified database, the data processing module adopts an NLP technology to perform structural processing on the acquired unstructured text data, extracts medicine names, indications, doses, pharmacological actions and compatibility tabulations, the model training module classifies and performs association analysis on prescription combinations based on a machine learning algorithm to generate model parameters for prescription matching, and the prescription matching module can enable users to perform a function according to purposes, the preparation, action mechanism, region and other screening conditions are searched. Although the most qualified prescription combination is selected from the database according to the matching algorithm and the training model, the traditional prescription combination which meets the conditions is recommended to the user through the condition screening and the matching algorithm, and the system is a system based on historical experience and data retrieval, is essentially to search for known prescription, and does not deeply explore why the prescription is effective and what the inherent biological action mechanism is. In the second prior art, publication number CN113486231a relates to a traditional Chinese medicine prescription and a disease analysis method, an analysis system, a device and a storage medium, and by combining compound information, a drug action target information set, a disease target information set and a preset target interaction matrix, a set of analysis mechanism for targets in the prescription-disease system is formed, analysis results are obtained, and ranking is performed on the analysis results, so that target information with the greatest degree of correlation is obtained. Although the accuracy and reliability of the network pharmacological research mechanism are improved, the network pharmacological research mechanism focuses on the relatively simple linear or static network relation of components, targets and diseases, the analysis core is to search for a common target and construct a network diagram based on the common target, the prescription is often regarded as a collection of components, and the depth analysis of how the components cooperate to influence the dynamic process of a disease network is lacking. In the prior art, publication number CN111241164A relates to a pharmacology analysis platform and analysis method of a traditional Chinese medicine system, and the pharmacology analysis platform comprises a data mining module, a data processing module, a target determination module, a Venturi image generation module