Computing Method For Herb Target Prediction

In view of the micro drug basic network characteristics, network pharmacology has become the new ideas of drug mechanism research and new drug research and development. Drug target prediction based on data and computational is one of the core research network pharmacology. In recent years, modern drug target prediction has been made much progress. However, due to the diversity and complexity of traditional Chinese medicine chemical composition, and lack of network pharmacology data net,the process of herb target predicting research which based computational has been make slow, no specialized analytical herb target predicting methods for characteristics of Traditional Chinese medicine. Based on the current access to data that relationship between the chemical composition of Traditional Chinese medicine,and herb target data and other resources, by building a complex network of traditional Chinese medicine-related targets, this paper conducted the Chinese target prediction analysis in the following three aspects:(1) According to the basic characteristics of multi-component of Traditional Chinese Medicine, integration of relevant Traditional Chinese Medicine data of chemical composition, chemical composition and protein relations, proposed a variety of drugs similarity computational method, which combines chemical structure and fingerprint of multi-component of Traditional Chinese Medicine, it’s able to calculate the similarity between medicine which has the characteristics of multi-component,and provide a basis for Traditional Chinese Medicine targets predictive analysis methods;(2) Proposed method of herb target prediction which basd on complex network,this method take advantage of the information of Traditional Chinese Medicine similarity and target protein network structure. Combined link prediction and machine learning methods, predictive analysis carried herb targets and achieved practical results;(3) Against the difficult problem of negative sample selected in machine learning-based herb target prediction,proposed a variety of method of negative sample selected, which the herb target prediction method is formed on the basis of, obtain a better prediction compared to the baseline performance. It’s shown to be an effective improvements.