Research For Statistical Analysis Of Drug Mode Of Action Based On Multi Features Fusion

Identifying drug mode of action(MoA)is of paramount importance for drug design and development,side effect prediction,and having a good grasp of drug indications in clinical tests.With the declining trend of new drug yield year by year,predicting drug MoAs has been recognized as one of the effective methods to solve the high input and low yield of drug development.Comprehensive knowledge of the MoAs of available drugs can help to identify possible new applications and unknown side effects.In this paper,we selected the US Food and Drug Administration(FDA)approved drugs as the research object.By statistical analysis and data mining methods,we analyze the relationship between the drug different information and MoAs,and mine the difference of drug properties among different drug MoAs.We use the probability ensemble method with integrated features to predict potential drug MoAs that can provide a reference basis for clinical treatment.The main contents are as follows:1.We proposed a method to analyze drug MoA through drug four features:chemical structure,biological property,pharmacological property and side effect.The method mainly uses these drug properties as feature to analyze difference and enrichment of drug information in various drug MoAs by statistical methods,and validates whether these features can distinguish different drug MoAs.The results show that there are significant differences in different drugs’ properties under different MoAs,and verify these drug information can be used to well distinguish the different drug MoAs.2.We proposed a probability ensemble method to predict drug MoAs based on multi feature fusion.This method integrated drug chemical structure,biological property,pharmacological property and side effect based on the Bayesian network theory,and use the probability theory to construct drug MoA network to find potential MoAs.The experimental results show that the multi feature fusion is effective in the performance of predict models and probability ensemble method has the higher prediction accuracy and better robustness than the other four methods.The model successfully predicts some potential drug modes of action.3.We constructed an online pharmacological database "MoABank" about drug MoA.The database can provide more comprehensive information about drug MoAs,drug targets,pathways,side effects and so on,as well as the analysis results based on the proposed model.