Active learning guided drug design lead optimization based on relative binding free energy modeling

• DOI: 10.26434/chemrxiv-2022-krs1t
• Publication Date: 2022-07-11T07:10:21Z
• Authors: Filipp Gusev; Evgeny Gutkin; Maria G. Kurnikova; Olexandr Isayev
• Keywords: Ligand efficiency; Free energy perturbation

In silico identification of potent protein inhibitors commonly requires prediction a ligand binding free energy (BFE). Thermodynamics integration (TI) based on molecular dynamics (MD) simulations is BFE calculation method capable predicting accurate BFE, but it computationally expensive and time-consuming. this work, we developed an efficient automated workflow for identifying compounds with the lowest among thousands congeneric ligands which only hundreds TI calculations. Automated Machine Learning (AutoML) orchestrated by Active (AL) in AL-AutoML allows unbiased search small set best performing molecules. We applied to select SARS-CoV-2 papain-like protease. Our work resulted 133 improved affinity 16 better than 100-fold improvement. The hit rate obtained here that traditional projects where molecule selection guided expert medicinal chemist. demonstrated combination AL protocol provides at least 20x common brute force approaches.