Predicting Potential Drug-Target Interactions with Multi-label Learning and Ensemble Learning

摘要

Identifying Drug-Target Interactions (DTIs) is an important process in drug discovery. Wet experimental methods arc expensive and time-consuming for detecting DTIs. Therefore, computational approaches were provided to deal with this task and have many effective strategies. In this paper, by using a machine-learning models for identifying drug targets, we provide a novel tool for DTIs prediction. In recent years, most of computational methods only are used to find the drug-drug similarity or target-target similarity, which cannot perfectly capture all characteristics to identify DTIs. To improve the performance of prediction, we focus on critical drug-related features and ignore irrelevant features to represent drugs, targets and relationship between them. Moreover, we further develop the ensemble learning model by integrating individual feature-based multi-label models for predicting DTIs. Experiments of evaluation show that the proposed approach achieves better results than other outstanding methods on benchmark datasets.