Predicting Nirmatrelvir Resistance in SARS-CoV-2 M^pro Mutants with an Integrated Computational Framework

Quantifying the thermodynamic impact of amino acid substitutions on the binding of nirmatrelvir to the SARS-CoV-2 main protease (M^pro) is fundamental for predicting drug resistance and informing the rational design of next-generation inhibitors. Here, we present an integrated computational pipeline to systematically map the resistance landscape of the M^pro-nirmatrelvir complex. Using alanine-scanning approach based on the generalized Born and interaction entropy (ASGBIE), 12 residues within 4 Å of the ligand were identified as key residues, including M49, Y54, N142, S144, H163, M165, E166, L167, P168, H172, Q189, and Q192. Subsequent saturation mutagenesis using different alchemical free-energy methods, including free-energy perturbation (FEP), thermodynamic integration (TI), multistate Bennett acceptance ratio (BAR/MBAR), and an averaged approach on these key residues, evaluates 228 single-point mutations. We found that M165 and L167 exhibit the lowest mutational tolerance, with most mutations at these sites compromising inhibitor affinity. Beyond identifying reported drug-resistant mutations such as E166 V/M and M165T, our large-scale scanning predicted several unreported high-risk mutations, including M165E/K and L167Y/W. Structural analysis shows that these mutations weaken binding through the loss of hydrophobic packing in the S2 and S3/S4 subsites and the disruption of the hydrogen-bonding network in the S1 pocket. These results reveal the residues most vulnerable to mutations, and our computational pipeline provides a reliable tool for predicting drug-resistance pathways and designing more robust M^pro inhibitors.