Research On Classification Of Family Of Traditional Chinese Medicine And Related Analysis Of Drug Properties Based On Complex Network

The combination of TCM theory and pharmacological research and modern advanced computer technology provides strong scientific support for the medicinal properties of Chinese medicine and helps people to further explore the laws of medicinal properties.The medicinal theory of traditional Chinese medicine includes the four properties,five flavors,functions and indications,meridian tropism,ascending,descending,and so on,which provides guidance for the rational use of drugs in clinical practice.Botanical medicine occupies a large proportion of traditional Chinese medicine and has the characteristics of abundant resources and low cost.Due to differences in plant growth environment and growing seasons,Chinese medicines have their own medicinal properties,which can be reflected by the morphological characteristics of plants.Existing research shows that there may be a connection between the family,physical appearance and medicinal properties of botanicals.However,the current research mainly adopts frequency analysis method,unable to find the potential relationship between family and pharmacological properties.This article aims to explore the network associations and potential molecular mechanisms of the medicinal families based on the information of the medicinal properties,families and molecules of botanicals,combined with information technology such as correlation analysis,network medical analysis,and machine learning,to form a predictive analysis method of medicinal properties based on plant morphology,mainly carried out the following three parts of research:(1)Through crawler technology,collect and integrate botanical-related basic source information,medicinal properties information(nature,flavour,meridian tropism and function),pharmacological information(ingredients and targets),forming data resources including Chinese medicine,family,medicinal properties,ingredients,and genes.Furthermore,by combining the network analysis method to carry out the network mechanism analysis of the family and medicinal properties of traditional Chinese medicine,we found the family and medicinal properties distribution law of botanicals,as well as the medicinal properties distribution law of each family of Chinese medicines:8% of the 500 botanicals Chinese medicines come from Compositae,6% of Chinese medicines come from legumes,59% of Chinese medicines are bitter,and 27% of Chinese medicines have the function of clearing away heat.(2)In view of the relationship between the medicinal properties of traditional Chinese medicine and the family mechanism,a framework for the analysis of the correlation mechanism between the family and family of drugs based on association rules,mutual information and statistical testing was established.Specifically,through association rules,mutual information and other association analysis methods,we found closely related medicinal properties-family and genus relationship pairs,such as:compositae and bitter,ginger family and pungent taste,ginger and spleen,rutaceae and qi-regulating,etc.Furthermore,an analysis of the consistency of the properties of Chinese medicines through euclidean distance in the same family was carried out,and the results showed that the consistency of the properties of most of the family was higher than that of random cases.Based on Fisher's exact test and statistical analysis method,the medicinal properties of traditional Chinese medicine and molecular mechanism analysis of families were carried out from the perspectives of ingredients and targets.The results showed that the medicinal properties of Chinese medicine acted through some ingredients and targets;Chinese medicines of the same family share some important ingredients and targets.(3)Aiming at the problem of classifying the properties of traditional Chinese medicines,a classification method based on the morphological characteristics of traditional Chinese medicines is proposed.Firstly,we crawled the pictures of the source plants corresponding to Chinese medicines from the Chinese Plant Image Database,and constructed the image data set of the base source plants of Chinese medicines.On this basis,a classification framework for medicinal properties(nature,flavor,meridian tropism)based on plant morphology is proposed.Using traditional machine learning and deep learning methods to classify the properties of traditional Chinese medicines,the experimental results show that the classification model based on Res Net18 has achieved good classification performance(F1 value can reach 0.91)in the tasks of nature,flavor and meridian tropism,indicating the medicinal properties of traditional Chinese medicine are potentially related to the morphological characteristics of the source plant.