Research On Prediction Of Drug-drug Interaction Based On Multi-feature Fusion

Drug-drug interaction refers to the mutual influence of two or more drugs existed simultaneously inside the body.Such mutual effects may lead to adverse events and serious drug safety problems,leading to the fact that the study of methods of predicting drug-drug interactions has become a hot topic in the field of drug research and is of great importance for drug development and clinical application.To illustrate,the prediction can guide clinical drug use,save drug development costs,and promote drug innovation and personalized treatment.Therefore,the paper focuses on the prediction of drug-drug interactions and conducts following studies:(1)The existing prediction methods are mainly based on the structural information of drug molecules,ignoring the multimodal information of drugs in the body.Therefore,the paper proposes a method of multimodal drug-drug interaction prediction based on hybrid fusion,which predicts the interaction through fusing data from multiple modalities improving the prediction accuracy and robustness.Firstly,the data of multiple modalities,including that of drug molecule structures,drug-target interactions and the drug metabolism,are collected and preprocessed,and the data’s features are extracted.Then,a hybrid fusion model was designed to fuse the features of different modalities.In this model,each component processes and learns separately features of the data in different modalities deriving the final prediction results.The experimental results show that the proposed method has higher accuracy compared with the traditional single-feature prediction methods and other multimodal methods.(2)The present prediction methods rely only on the statistical characteristics of data,while ignoring the uncertainty and incompleteness of data,which may lead to inaccurate and misleading predictions.Therefore,a drug-drug interaction prediction method based on evidence theory is proposed.The method transforms the problem of predicting multimodal drug interactions into a combination problem of evidence and then assesses the confidence level of prediction by calculating the weight and credibility of each evidence.Specifically,the method converts data of different modalities into different evidences and models these evidences using the Dirichlet distribution,and then integrates and reasons about these evidences using an evidence theory to produce the final prediction results.In this process,the uncertainty of different evidences is considered.The experimental results show that the method can make full use of data of modalities to improve prediction accuracy and robustness,and also has strong interpretability.The study of drug-drug interaction prediction methods is an important topic.Based on the review and analysis of existing research results,this paper provides an in-depth discussion on the application of multimodal fusion and evidence theory in drug-drug interaction prediction,which provides new ideas and methods to further improve the accuracy and credibility of interaction prediction.