(442f) Combating COVID-19 Using High Resolution Computational Protein Docking with Natural Drug Products

SARS-CoV-2 is the virus responsible for the 2020 global pandemic. Although multiple vaccines have been approved for use against the virus throughout the world, the threat of disease remains largely unsolved for individuals with underlying medical conditions and people residing in regions where vaccines and the necessary drug treatments are not available. A potential solution to this problem is the use of phytochemicals (chemical compounds obtained from plants) that inhibit viral activity through the binding of certain SARS-CoV-2 proteins. The phytochemicals selected in this study were obtained from the USDA Phytochemical and Ethnobotanical Database, and many of them have documented anti-viral properties and are generally regarded as safe. To identify potential therapeutic phytochemicals, we used the Rosetta software, a computational biology tool with ligand-protein docking capabilities, to screen a phytochemical library against selected viral proteins. From this analysis, we identified 19 potential binders for the SARS-CoV-2 main protease and 16 potential binders for various non-structural proteins. (e.g., NSP3, NSP9 RNA-replicase, and NSP16-NSP10 complex) In addition to conducting the screening, we also investigated how the Rosetta docking performance was affected by different score functions, ligand protonation states, and the number of Metropolis Monte Carlo cycles. To validate our docking results, we performed correlation tests between the strengths of the computationally predicted protein-phytochemical interactions and experimentally obtained protein-ligand binding affinities. Further validation tests performed between the computationally predicted atomic positions and experimental crystal structure atomic positions highlight the current limitations and challenges of in silico docking. To improve the predictive power of our docking workflow, we incorporate multiple distinct ligand-protein docking protocols validated against an expanded collection of experimentally derived structural and biophysical data. Our follow-up studies will apply supervised machine learning models to relate the structural and energetic data of thousands of decoys (computationally obtained protein-ligand structures) and experimentally obtained crystal structures in order to better predict protein-ligand docking. Overall, this study will provide useful information on plant-based drug development for COVID-19 treatment. The models and high throughput docking protocol established may be used in a wide range of protein-ligand docking scenarios.