SUMMARYAlthough numerous promising therapeutic targets for human diseases have been discovered, most have not been successfully translated into clinical practice1. A bottleneck in the application of basic research findings to patients is the enormous cost, time, and effort required for high-throughput screening of potential drugs2 for given therapeutic targets. Recent advances in 3D docking simulations have not solved this problem, given that 3D protein structures with sufficient resolution are not always available and that they are computationally expensive to obtain. Here we have developed LIGHTHOUSE, a graph-based deep learning approach for discovery of the hidden principles underlying the association of small-molecule compounds with target proteins, and we present its validation by identifying potential therapeutic compounds for various human diseases. Without any 3D structural information for proteins or chemicals, LIGHTHOUSE estimates protein-compound scores that incorporate known evolutionary relations and available experimental data. It identified novel therapeutics for cancer, lifestyle-related disease, and bacterial infection. Moreover, LIGHTHOUSE predicted ethoxzolamide as a therapeutic for coronavirus disease 2019 (COVID-19), and this agent was indeed effective against alpha, beta, gamma, and delta variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that are rampant worldwide. Given that ethoxzolamide is already approved for several diseases, it could be rapidly deployed for the treatment of patients with COVID-19. We envision that LIGHTHOUSE will bring about a paradigm shift in translational medicine, providing a bridge from bench side to bedside.
Reinforcement Learning to Boost Molecular Docking upon Protein Conformational Ensemble