Graph Convolutional Networks for Predicting Drug-Protein Interactions

In this paper, a heterogeneous graph of drug-target entities are constructed to predict drug-protein interactions. We use a deep learning approach and apply an encoder-decoder technique in an end-to-end manner directly on a full-scale heterogeneous graph. The proposed method not only achieves performance improvement over previous state-of-the-art techniques, but also integrates additional information into the model for a more comprehensive analysis.