The COVID-19 pandemic prompted several drug repositioning initiatives with the aim to rapidly deliver pharmacological candidates able to reduce SARS-CoV-2 dissemination and mortality. A major issue shared by many of the in silico studies addressing the discovery of compounds or drugs targeting SARS-CoV-2 molecules is that they lacked experimental validation of the results. Here we present a computer-aided drug-repositioning campaign against the indispensable SARS-CoV-2 main protease (MPro or 3CLPro) that involved the development of ligand-based ensemble models and the experimental testing of a small subset of the identified hits. The search method explored random subspaces of molecular descriptors to obtain linear classifiers. The best models were then combined by selective ensemble learning to improve their predictive power. Both the individual models and the ensembles were validated by retrospective screening, and later used to screen the DrugBank, Drug Repurposing Hub and Sweetlead libraries for potential inhibitors of MPro. From the 4 in silico hits assayed, atpenin and tinostamustine inhibited MPro (IC50 1 µM and 4 μM, respectively) but not the papain-like protease of SARS-CoV-2 (drugs tested at 25 μM). Preliminary kinetic characterization suggests that tinostamustine and atpenin inhibit MPro by an irreversible and acompetitive mechanisms, respectively. Both drugs failed to inhibit the proliferation of SARS-CoV-2 in VERO cells. The virtual screening method reported here may be a powerful tool to further extent the identification of novel MPro inhibitors. Furthermore, the confirmed MPro hits may be subjected to optimization or retrospective search strategies to improve their molecular target and anti-viral potency.

No abstract