Deep Learning Can Identify Explainable Reasoning Paths of Mechanism of Drug Action for Drug Repurposing from Multilayer Biological Network

The discovery and repurposing of drugs require a deep understanding of the mechanism of drug action (MODA). Existing computational methods model these mechanisms with the help of protein-protein interaction (PPI) network. However, the molecular interactions of drugs in the human body are far beyond the PPI network, and the lack of interpretability of these models hinders their practical applications. In this study, we propose iDPath, an interpretable deep learning-based path-reasoning framework to identify potential drugs for the treatment of diseases by capturing the MODA based on simulating the paths from drugs to diseases in the human body. We create a multilayer biological network consisting of a gene regulatory layer, a PPI layer, a protein-chemical interaction layer, and a chemical-chemical interaction layer to comprehensively characterize the molecular interactions in the human body. Based on this multilayer biological network, iDPath utilizes a graph convolutional network module to capture the global connectivity information of the human molecular network; and a long short-term memory neural network module to capture the detailed mechanisms of drug action by modeling the sequential dependencies of the MODA-related biological paths between drugs and diseases. Moreover, iDPath applies two attention modules - node attention and path attention - to further enhance its interpretability. Experiments with drug screen data show that iDPath outperforms state-of-the-art machine learning methods on a general drug repurposing task featuring 1,993 drugs and 2,794 diseases. Further investigations demonstrate that iDPath can identify explicit critical paths that are consistent with clinical evidence. To demonstrate the practical value of iDPath, we apply it to identify potential drugs for the treatment of prostate cancer. Results show that iDPath can successfully discover new FDA-approved drugs. This research provides a novel interpretable artificial intelligence perspective on drug discovery.