Using Novel Convolutional Neural Networks Architecture to Predict Drug-Target Interactions

Identifying potential drug-target interactions (DTIs) are crucial task for drug discovery and effective drug development. In order to address the issue, various computational methods have been widely used in drug-target interaction prediction. In this paper, we proposed a novel deep learning-based method to predict DTIs, which involved the convolutional neural networks (CNNs) to train a model and yielded robust and reliable predictions. The method achieved the accuracies of 92.0%, 90.0%, 92.0% and 90.7% on enzymes, ion channels, GPCRs and nuclear receptors in our curated dataset, respectively. The experimental results indicated that our methods improved the DTIs predictions in comparison with the state-of-the-art computational methods on the common benchmark dataset.