X-DPI: A structure-aware multi-modal deep learning model for drug-protein interactions prediction

Motivation Identifying the drug-protein interactions (DPIs) is crucial in drug discovery, and a number of machine learning methods have been developed to predict DPIs. Existing methods usually use unrealistic datasets with hidden bias, which will limit the accuracy of virtual screening methods. Meanwhile, most DPIs prediction methods pay more attention to molecular representation but lack effective research on protein representation and high-level associations between different instances. To this end, we presented here a novel structure-aware multi-modal DPIs prediction model, X-DPI, performing on a curated industry-scale benchmark dataset. Results We built a high-quality benchmark dataset named GalaxyDB for DPIs prediction. This industry-scale dataset along with an unbiased training procedure resulted in a more robust benchmark study. For informative protein representation, we constructed a structure-aware graph neural network method from the protein sequence by combining predicted contact maps and graph neural networks. Through further integration of structure-based representation and high-level pre-trained embeddings for molecules and proteins, our model captured more effectively the feature representation of the interactions between them. As a result, X-DPI outperformed state-of-the-art DPIs prediction methods and obtained 5.30% Mean Square Error (MSE) improved in the DAVIS dataset and 8.89% area under the curve (AUC) improved in GalaxyDB dataset. Moreover, our model is an interpretable model with the transformer-based interaction mechanism, which can accurately reveal the binding sites between molecule and protein.