High-throughput screening (HTS) assays have been developed and applied to identify potential lead compounds for COVID-19 drug development. However, traditional HTS of large compound libraries often suffer from high costs and low hit rates as well as limitation of available compound collections in house. Because SARS-CoV-2 is a biosafety level 3 (BLS-3) virus, assays using live viruses have to be containment facilities. To address these limitations, we have developed machine learning models for virtual screens to identify potential anti-SARS-CoV-2 compounds acting via two different mechanisms: inhibition of the entry of SARS-CoV-2 into human host cells or the SARS-CoV-2 3-chymotrypsin-like (3CL) protease. The optimal classification models achieved good performance with the area under the receiver operating curve (AUC-ROC) values &gt; 0.78. The compounds identified from the virtual screens using these models were confirmed experimentally. The best performing models increased the assay hit rate by 2.1-fold for viral entry inhibitors and 10.4-fold for 3CL protease inhibitors compared to the original drug repurposing screens. The anti-SARS-CoV-2 activities of the experimentally confirmed inhibitors were further validated in a SARS-CoV-2 live virus assay that measured the cytopathic effect of viral infection. Twenty-two compounds showed potent (&lt; 5 µM) anti-SARS-CoV-2 activities that have the potential to be developed as novel COVID-19 drug leads. In conclusion, machine learning models can be developed and used as a complementary approach to HTS to expand compound screening capacities and improve the speed and efficiency of anti-SARS-CoV-2 drug discovery.

Efficient identification of anti-SARS-CoV-2 compounds using machine learning

ACS Spring 2022

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