Computationally guided side chain optimization of nirmatrelvir for lead discovery against coronavirus main proteases

To optimize Nirmatrelvir as an inhibitor of the MERS-CoV main protease (Mpro), we developed an automated RDKit-based iterative scaffold optimization framework that integrated fragment replacement, combinatorial library enumeration, and chemical feasibility filtering to generate 15,744 drug-like analogs through systematic substitutions at six positions on the Nirmatrelvir scaffold. Structural similarity to Nirmatrelvir was assessed using complementary 2D fingerprint and 3D shape-based analyses of the 15,953 designed analogs and a separate set of 13,090 molecular weight-matched ChEMBL compounds to examine the binding potency in both the modified and randomly selected compound cohorts. Based on the correlation between structural similarity and docking score of the modified molecules, a binding score (BS) approach was advanced to rapidly identify and rank promising lead compounds. Several top candidates exhibited improved binding characteristics compared to that of Nirmatrelvir. To assess the robustness of the BS approach, multiple computational methods were employed, including K DEEP , molecular dynamics (MD) simulations, molecular mechanics/generalized Born surface area (MM/GBSA) free energy calculations, and Boltz-2. The BS ranking showed strong agreement with all computational validation methods, consistently identifying compounds with enhanced binding affinities and stable protein-ligand interactions. The top-ranked compounds, M1, M3, M5, and M6, identified by high Binding Score (BS) values (9.1–10.2), exhibited favorable MM/GBSA binding free energies ranging from − 31.7 to − 42.9 kcal/mol, K DEEP -predicted binding affinities of − 9.5 to − 10.6 kcal/mol, and Boltz-2 predicted pIC₅₀ values of 10.0–11.8. Furthermore, for SARS-CoV-2 Mpro, BS scores also correlated well with available experimental half-maximal inhibitory concentration (IC 50 ) data and Boltz-2 predictions for Nirmatrelvir analogues. Collectively, these findings demonstrate that the BS approach may serve as an efficient strategy for screening, prioritizing, and accelerating the discovery of next-generation coronavirus Mpro inhibitors.