Research On The Molecular Characteristics Of Chinese Herbal Medicine With Different Properties Based On Data Mining

The property theory of Chinese Herbal Medicine(CHM)is the foundation and core of the theoretical system of Traditional Chinese medicine(TCM)and guides the clinical use of CHM.However,this traditional theory is constantly challenged by modern medicine because of its empirical,abstract,fuzzy and other characteristics.Therefore,based on the chemical components of CHM with different medicinal properties,this paper used modern scientific methods to analyze and study the properties of CHM.The main work of this paper is as follows:Firstly,the molecular data of CHM were preprocessed.This paper proposed a molecular vector model to characterize CHM which can fully consider the compounds contained in each CHM.The MACCS molecular fingerprint and molecular descriptors were used to describe the structure and properties of the molecules.Secondly,the classification model of CHM properties was established based on data mining classification algorithm.In this paper,six classification algorithms,including logistic regression,decision tree,Ada Boost,XGBoost,naive Bayes and random forest,were selected as the bottom classifiers.The voting fusion strategy was used to integrate the six models,and the accuracy of different model combinations on the test set was compared.The results show that logistic regression,naive bayes and random forest after Voting fusion had the highest prediction accuracy,which was 93%.Experiments show that the model has a certain reliability,and can provide auxiliary reference for the classification of unrecognized CHM.Finally,based on the improved atomic association rule algorithm and agglomerative hierarchical clustering algorithm,the differences of molecular characteristics of CHM with cold and hot properties were studied.Experimental results show that CHM with the same property have certain commonalities in compound composition.The key molecules related to the cold property of CHM were mainly polycyclic compounds,and the key molecules related to the hot property of CHM were mainly monocyclic or acyclic compounds.There were significant differences in molecular weight,number of oxygen atoms,number of rings,number of aromatic bonds,total information index of atomic composition,topological polarity surface area and Log P value of the key molecules of cold and hot properties.This part of the study provides new ideas and methods for understanding and explaining the properties of CHM.