Background/Objectives: The EGFR T790M gatekeeper mutation drives resistance in non-small cell lung cancer by sterically hindering inhibitors and restoring ATP affinity. As emerging C797S mutations render covalent inhibitors obsolete, novel non-covalent strategies are critical. This study identifies inhibitors that redefine the mutant methionine sulfur atom as a primary stabilizing anchor rather than a liability. Methods: A reinforcement learning-based generative AI framework (DrugEx) sampled 100,000 molecules, prioritized through QSAR-based classification (mean ROC-AUC: 0.91 ± 0.01) and physicochemical filtering. Four lead candidates underwent 200 ns all-atom molecular dynamics (MD) simulations. Mechanistic stability and energetic convergence were quantified via Free Energy Landscape (FEL) analysis, post-simulation Ramachandran validation, and ensemble-averaged Molecular Mechanics Generalized Born Surface Area (MM-GBSA) binding free energy calculations. Results: Candidate 106 demonstrated high mutation tolerance by redistributing interactions toward the Met790 sulfur atom, contrasting with reference scaffolds 14 and 88 which suffered significant affinity loss. MD analysis revealed that potency is fundamentally dictated by successful recruitment of the thioether environment, locking the complex within a narrow, well-defined thermodynamic basin. Candidate 106 maintained stable binding (−11.0 kcal/mol by docking) corroborated by a robust MM-GBSA ΔGbind of −50.51 kcal/mol, primarily driven by persistent π-sulfur contacts (85% occupancy). Ramachandran analysis confirmed over 90% favored region occupancy, ensuring interactions occur without non-physical protein strain. Conclusions: Successful T790M/C797S resistance bypass is achievable by exploiting the gatekeeper methionine’s electronic environment. This mutation-aware blueprint provides a non-covalent strategy that avoids the metabolic and resistance-prone liabilities of covalent warheads.

Background/Objectives: The EGFR T790M mutation drives lung cancer resistance by sterically hindering inhibitors and restoring ATP affinity. As C797S mutations render covalent inhibitors obsolete, novel non-covalent strategies are critical. This study identifies inhibitors that redefine the mutant methionine sulfur as a primary stabilizing anchor rather than a liability. Methods: A generative AI framework (DrugEx) sampled 100,000 molecules, prioritized via QSAR classification (ROC-AUC: 0.91 ± 0.01) and Applicability Domain (AD) mapping. The workflow was de-risked through retrospective benchmarking against the DUD-E database (35,590 molecules), achieving a 1% Enrichment Factor of 5.19. Lead candidates underwent 100 ns all-atom molecular dynamics (MD) simulations. Mechanistic stability was quantified via Free Energy Landscape (FEL) analysis and ensemble-averaged MM-GBSA binding free energy calculations. Results: Candidate 106 demonstrated exceptional mutation tolerance by redistributing interactions toward the Met790 sulfur atom. MD analysis confirmed potency is dictated by successful recruitment of the thioether environment, locking the complex within a narrow thermodynamic basin. Candidate 106 maintained stable binding (−11.0 kcal/mol) corroborated by an equipotent MM-GBSA ΔG_bind of −50.51 kcal/mol in the mutant system, driven by persistent π-sulfur contacts (85% occupancy). Conclusions: These results indicates that potential T790M resistance bypass is achievable by exploiting the gatekeeper methionine’s electronic environment. This modeled mutation-aware framework provides a candidate non-covalent strategy to be validated in future wet-lab campaigns.