Drug-Target Interaction Prediction Based on Heterogeneous Networks

Predicting drug-target interactions has gradually become a heated issue in medical research. However, identifying the drug-target interactions in clinical trials requires a lot of financial resources and time. More and more computational methods are currently used for drug-target interaction predictions. This paper proposes a drug-target interaction prediction method that can integrate information from different heterogeneous networks. This method constructs multiple drug and target similarity networks and applies the GraRep algorithm on the similarity networks after denoising in order to extract the features. Features obtained from heterogeneous networks are combined as a feature vector of DTIs, which is the input of Random Forest. The result of our experiment shows that our method in this paper increases the accuracy of prediction for DTIs, which is superior to other state-of-the-art approaches.