The current outbreak of COVID-19 is leading an unprecedented scientific effort focusing on targeting SARS-CoV-2 proteins critical for its viral replication. Herein, we performed high-throughput virtual screening of more than eleven thousand FDA-approved drugs using backpropagation-based artificial neural networks (q2LOO = 0.60, r2 = 0.80 and r2pred = 0.91), partial-least-square (PLS) regression (q2LOO = 0.83, r2 = 0.62 and r2pred = 0.70) and sequential minimal optimization (SMO) regression (q2LOO = 0.70, r2 = 0.80 and r2pred = 0.89). We simulated the stability of Acarbose-derived hexasaccharide, Naratriptan, Peramivir, Dihydrostreptomycin, Enviomycin, Rolitetracycline, Viomycin, Angiotensin II, Angiotensin 1-7, Angiotensinamide, Fenoterol, Zanamivir, Laninamivir and Laninamivir octanoate with 3CLpro by 100 ns and calculated binding free energy using molecular mechanics combined with Poisson-Boltzmann surface area (MM-PBSA). Our QSAR models and molecular dynamics data suggest that seven repurposed-drug candidates such as Acarbose-derived Hexasaccharide, Angiotensinamide, Dihydrostreptomycin, Enviomycin, Fenoterol, Naratriptan and Viomycin are potential SARS-CoV-2 main protease inhibitors. In addition, our QSAR models and molecular dynamics simulations revealed that His41, Asn142, Cys145, Glu166 and Gln189 are potential pharmacophoric centers for 3CLpro inhibitors. Glu166 is a potential pharmacophore for drug design and inhibitors that interact with this residue may be critical to avoid dimerization of 3CLpro. Our results will contribute to future investigations of novel chemical scaffolds and the discovery of novel hits in high-throughput screening as potential anti-SARS-CoV-2 properties.Communicated by Ramaswamy H. Sarma.
Keywords: QSAR models; SARS-CoV-2; high-throughput virtual screening; molecular dynamics simulations; pKi prediction.