Drug-Target Interaction Prediction with Weighted Bayesian Ranking

Identifying drug-target interactions (DTIs) through biochemical experiments is very expensive and time-consuming. Therefore, it is an inevitable trend to use computational methods to predict the drug-target interactions, and high prediction accuracy becomes our ultimate goal. However, most existing computational methods treat the non-interaction data as negative samples which is unreasonable as those non-interaction data may contain undetected drug-target interactions. In this paper, a novel weighted Bayesian ranking method (WBRDTI) for drug-target interactions prediction is proposed, and the different effects of each drug-target pair also is taken into account. Besides, dual similarity is used to regularize the latent factors of drugs and targets respectively, and known neighbor information is used to smooth novel drug or target. Finally, the experiment results on widely used publicly available drug-target interaction datasets show its effectiveness and the practicality of the proposed method.