DeepDrug: A general graph-based deep learning framework for drug-drug interactions and drug-target interactions prediction

Exploring the biomedical interactions about chemical compounds and protein targets is crucial for drug discovery. Determining these interactions (DDI/DTI) not only reveals the potential synergistic effects of drug combinations and improves drug efficacy, but also contributes to drug reuse, reduces drug development costs, and improves drug development efficiency. Therefore, predicting interactions among drugs and drug targets is an important topic in the field of drug discovery.

Recently, Quantitative Biology published an approach entitled “DeepDrug: A general graph-based deep learning framework for drug-drug interactions and drug-target interactions prediction”, which shows DeepDrug learns the comprehensive structure- and sequence-based representations of drugs and proteins, achieving optimal performance across a range of tasks, by leveraging the residual graph convolutional networks and convolutional networks.

DeepDrug (Figure 1) predicts drug/target interactions by combining sequence features and structural features, leveraging convolutional module and residual graph convolutional submodules, respectively. DeepDrug outperforms state-of-the-art methods in a series of systematic experiments, including binary-class DDIs, multi-class/multi-label DDIs, binary-class DTIs classification and DTIs regression tasks. Furthermore, the structural features learned by DeepDrug displays compatible and accordant patterns in chemical properties and drug categories, providing additional evidence to support the strong predictive capabilities of DeepDrug. As an application, DeepDrug is applied to discover the potential drug candidates against SARS-CoV-2, where 7 out of 10 top-ranked drugs are reported in the relevant literature.