Computational Discovery of <scp>MERS</scp> ‐ <scp>CoV</scp> Main Protease Inhibitors Through Screening and Molecular Dynamics Simulations

Targeting the main protease (Mpro) of coronaviruses has emerged as a promising therapeutic strategy for combating viral infections. Despite the global health threat posed by Middle East Respiratory Syndrome Coronavirus (MERS-CoV), no vaccines or antiviral drugs have been approved to date for its treatment. With a mortality rate approaching 35%, MERS-CoV remains a critical concern, particularly due to its potential for increased transmissibility through mutation. The viral main protease plays a pivotal role in the proteolytic processing of viral polyproteins, making it an attractive target for antiviral drug development. In this study, an in silico high-throughput screening was performed to identify potential inhibitors of MERS-CoV Mpro. A compound library comprising small molecules was curated from diverse sources, including DrugBank, CHEMBL, and known protease inhibitors from the Protein Data Bank. Top candidates were selected using molecular docking combined with a similarity-based search strategy, which prioritized compounds known to interact with Mpro and predicted to exhibit high binding affinity at its active site. The top-ranking candidates were further evaluated through molecular dynamics (MD) simulations to assess the conformational stability of the ligand-protein complexes. Binding free energies were subsequently calculated using multiple computational approaches, including the deep learning-based K_DEEP model, molecular mechanics/generalized born surface area (MM/GBSA), and free energy perturbation (FEP). Among the screened compounds, two molecules X2A and DB11779 (Danoprevir) consistently demonstrated superior binding affinities and stable interactions with MERS-CoV Mpro. These results agree well with experimental equilibrium dissociation constant (K_D) and half-maximal inhibitory concentrations (IC50). These findings highlight the capability of modern computational methods to generate accurate and robust binding data.