Predicting drug–target interaction using positive-unlabeled learning

Identifying interactions between drug compounds and target proteins is an important process in drug discovery. It is time-consuming and expensive to determine interactions between drug compounds and target proteins with experimental methods. The computational methods provide an effective strategy to address this issue. The difficulties of drug–target interaction identification include the lack of known drug–target association and no experimentally verified negative samples. In this work, we present a method, called PUDT, to predict drug–target interactions. Instead of treating unknown interactions as negative samples, we set it as unlabeled samples. We use three strategies (Random walk with restarts, KNN and heat kernel diffusion) to part unlabeled samples into two groups: reliable negative samples (RN) and likely negative samples (LN) based on target similarity information. Then, majority voting method is used to aggregate these strategies to decide the final label of unlabeled samples. Finally, weighted support vector machine is employed to build a classifier. Four datasets (enzyme, ion channel, GPCR and nuclear receptor) are used to evaluate the performance of our method. The results demonstrate that the performance of our method is comparable or better than recent state-of-the-art approaches.