De novo design of potential SARS-CoV-2 main protease inhibitors using artificial intelligence and molecular modeling technologies

Abstract

De novo design of 95 775 potential ligands of SARS-CoV-2 main protease (Mpro), playing an important role in the process of virus replication, was carried out using a deep learning generative neural network that was developed previously based on artificial intelligence technologies. Molecular docking and molecular dynamics methods were used to evaluate the binding affinity of these molecules to the catalytic site of the enzyme. As a result, 7 leading compounds exhibiting Gibbs free energy low values comparable with the values obtained using an identical computational protocol for two potent non-covalent SARS-CoV-2 Mpro inhibitors used in calculations as a positive control were selected. The results obtained indicate the promise of applying identified compounds for development of new antiviral drugs able to inhibit the catalytic activity of SARSCoV-2 Mpro.