Research And Implementation Of Traditional Chinese Medicine Prescription Repositioning Method Based On Heterogeneous Network Representation Learning

The repositioning of traditional Chinese medicines(TCM)prescriptions is a strategy to explore new indications of TCM prescriptions,thereby expanding their scope of application and uses.Drug repositioning has attracted the attention of many researchers due to its advantages of low wastage rate,low cost,and short time.Although there has been a lot of research work on the repositioning of western medicines,the western medicine repositioning method cannot be directly applied to the repositioning of traditional Chinese medicines prescriptions due to the characteristics of "multi-component,multi-target and multi-path regulation" of Chinese medicine prescriptions.In addition,although some scholars have used network pharmacology to study the relocation of TCM prescriptions,they only use some tools to manually explore the specific disease association of a certain prescription,and no general computational method for predicting the relocation of TCM prescriptions has been found.Therefore,based on the concept of network pharmacology,this thesis integrates relevant Chinese medicine databases and literature to construct a heterogeneous information network of traditional Chinese medicine,combines heterogeneous network representation learning and link prediction methods,and deeply explores the relationship between traditional Chinese medicine prescriptions and diseases,the research proposes an end-to-end method for Chinese medicine prescription relocation based on heterogeneous network representation learning,and designs and implements a Chinese medicine prescription relocation prediction tool based on the proposed method.The main research work of the thesis is as follows:(1)The thesis innovatively transforms the problem of Chinese medicine prescription relocation into the problem of predicting the association between Chinese medicine prescriptions and diseases,and proposes a Chinese medicine prescription relocation method combining attribute information and heterogeneous network representation learning(Attributed-Aware Heterogeneous Network Pharmacology Learning,AHNPL).The AHNPL method can use Transformer technology to learn the prescription,disease attribute information and its topological neighborhood representation from the heterogeneous network of traditional Chinese medicine prescriptions without setting the meta-path,and extract high-level structure and semantic relations to improve the ability to predict the potential prescription-disease association.performance.Experimental results show that the method achieves the best performance compared to other benchmark drug relocalization methods and heterogeneous network representation methods.(2)Further,based on the research on the characteristics of the overall effect of traditional Chinese medicine prescriptions,this thesis uses the method of defining the prescription community and combining the community attention mechanism to summarize and explore the mechanism of action that is different from the traditional drug repositioning on the basis of the AHNPL method.This study proposes a traditional Chinese medicine prescription relocation method based on community attention mechanism enhancement(Community Attention Mechanism Enhanced AHNPL,CAHNPL).The CAHNPL method considers the traditional Chinese medicine prescriptions as a whole,and can obtain the feature representation of nodes based on different prescription communities,which is beneficial to enhance the ability to mine the potential associated semantics of prescriptions and diseases.Further experimental results show that CAHNPL has better performance on the TCM prescription relocation task.(3)Based on the above research results,by analyzing the relevant requirements,the thesis designs and implements a Web tool for relocation prediction of traditional Chinese medicine prescriptions-(AMTCM-Repos).Based on the Flask framework and the MVC architecture with the front-end and back-end separated,the tool implements functions such as prediction and recommendation related to the relocation of traditional Chinese medicine prescriptions and viewing interpretability paths.