(360ag) Identification of Potential TMPRSS2 Inhibitors By Virtual Screening Using Molecular Docking and Machine Learning

Repeated viral outbreaks in the past and the recent COVID-19 pandemic have highlighted the urgent necessity for a broad-spectrum antiviral. Host proteases implicated in the infection mechanism of multiple viruses are ideal targets for such drugs. The human Transmembrane Protease Serine Type 2 (TMPRSS2), which is used by several coronaviruses including SARS-CoV-2, has emerged as a favorable target for treating coronaviruses. In this work, we perform virtual screening of small molecule protease inhibitors from the MEROPS database by molecular docking with UCSF DOCK to identify potential TMPRSS2 inhibitors. We also describe our homology model of the protease used for docking in the early days of the pandemic, which was found to be in close agreement with the subsequently resolved crystal structure. Compounds with the highest docking scores, which included approved anticoagulants like idraparinux and fondaparinux, were then analyzed for key interactions with TMPRSS2. In addition, we implement machine learning models, which use the results from docking to identify potentially important interactions and suggest modifications to improve the selected inhibitors. We continue to monitor emerging variants and new studies to ensure that our strategy remains pertinent.