Predicting drug-drug interaction events base on deep neural network

Drug-drug interaction is one of the important issue related to health care systems. In pharmaceutical research, drug-drug interactions (DDI) can be a severe issue. There is a major need for a system to foresee the interactivity between drugs and to accomplish this need, so various algorithms of machine learning have been introduced. According to the studies, DDIs can result in several different events, predicting drug-drug interaction-relevant event can be beneficial, leading to drug toxicity or detrimental reactions caused by combined used drug. In this paper, we introduced DDICNN to predict most occurrence of DDIs.The results of this study were to collect DDI events achieved from the DrugBank dataset and extract ۶۵ DDI classification using affiliation analysis and event modification. A deep learning approach was proposed in which known as DDICNN, in which various drug characteristic are combined with deep learning to generate a predictive model based on the interactions among drugs. To acquire cross-modality representations of drug-drug couples and anticipate DDI incidents, DDICNN sets up a deep neural network by investigating ۴ types of drug features: targets, chemical substructures, pathways, and enzymes. Based on the results of computational experiments, DDICNN is highly accurate and efficient. The DDICNN method also outperforms other methods for DDI event prediction and baselining. It seems that chemical substructures are the most important of all components of drugs. The accuracy of the DDICNN arrives at ۰.۸۹۷ when all features are combined.

کلیدواژه ها:

Drug-drug interaction ، Drug repurposing ، Deep neural network