Predicting Drug-Target Interaction Using Deep Matrix Factorization

Abstract

In silico prediction of drug-target interaction can help to speed up the process of identifying unknown interactions between drugs and target proteins in pharmaceutical research. In this paper, we first exploit k-nearest neighbor technique to identify the reliable negatives (non-interacting pairs) among unlabeled data. Then, we employ a Deep Matrix Factorization to predict drug-target interaction to reveal the non-linearity relations among interacting drugs and targets. We evaluate the results using area under the curve metrics. Our approach is applied to public-domain benchmarks and compared against the state-of-the-art methods.