Designing Potential Inhibitors of SARS-CoV-2 Mpro Using Deep Learning and Steered Molecular Dynamic Simulations

Abstract

The COVID-19 pandemic raised an unprecedented race in biotechnology in search for effective therapies and a preventive vaccine. Scientists worldwide have been attempting to stop the viral infection by interfering with the biological function of the SARS-CoV-2 main protease (Mpro), a critical protein required for viral transcription and replication during infection. In this study, we employed an effective approach integrating deep learning model calculations and steered molecular dynamic simulations to generate highly promising inhibitors of SARS-CoV-2 Mpro. First, using deep learning calculations, a natural molecule that was identified as a potential inhibitor of SARS-CoV-2 Mpro was chemically altered to boost its ligand-binding affinity to the Mpro protease. The proposed compounds were then verified using steered molecular dynamic simulations to estimate their binding free energies to SARS-CoV-2 Mpro. The procedure was repeated until the binding free energies of the proposed compounds did not improve further. Overall, one proposed compound was shown to have a high nanomolar affinity, and two others were estimated to possess nanomolar affinities for SARS-CoV-2 Mpro, indicating that they are highly promising inhibitors of the protease. Absorption, distribution, metabolism, and excretion and toxicity analysis show that all three chemicals are drug-like compounds following the MACCS-II Drug Data Report database, orally absorbed, tightly attached to the plasma membrane, and noncarcinogenic in rats. The results obtained potentially support COVID-19 treatment.