In an effort to respond to the COVID-19 pandemic, Lawrence Livermore National Laboratory used machine-learning and mechanistic simulations to discover antiviral drug leads for SARS-CoV-2. The viral spike receptor binding domain (RBD) and main protease (MPro) were used as protein targets. A number of screening methods were used to discover small molecule hits, including molecular docking, a 3-D structure-based deep learning model, MM/GBSA free energy calculations, neural network GBSA score predictors, and molecular dynamics simulations. Thousands of compounds were purchased and tested using in-vitro assays for binding to spike protein, inhibition of MPro function and cell infection by spike-expressing pseudovirus. Compounds that inhibited the pseudovirus were further tested against live SARS-CoV-2 virus. Initial focus was on available world drugs and commercially available compound libraries, and later moved to synthesizing novel compounds. Of the screening methods used here, the best predictor of experimental results was MM/GBSA free energy; however, MM/GBSA scores computed by neural network models from chemical descriptor features did almost as well. In the pseudovirus assay, imatinib, lapatinib and 5-nitro-2,4-di-1-piperidinylbenzaldehyde [4-[(3-chloro-4-methylphenyl)amino]-6-(4-morpholinyl)-1,3,5-triazin-2-yl]hydrazone had IC50 values below 10mM, while candesartan cilexetil had an IC50 value of approximately ~88 mM against Mpro in a FRET-based activity assay . Comparatively, candesartan cilexetil had the highest selectivity index of all four compounds as its half-maximal cytotoxicity concentration 50 (CC50) value was the only one greater than the limit of the assay (&gt;100mM). These lead compounds are planned to be used as input to a generative molecular design pipeline to create additional novel compounds with better safety profiles.

Using machine-learning and mechanistic simulations to discover drug leads for SARS-CoV-2

ACS Spring Meeting 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.