Accelerating Scaffold Hopping in Fourth-Generation Epidermal Growth Factor Receptor Inhibitors via Multilevel Virtual Screening

The epidermal growth factor receptor (EGFR) family plays a crucial role in regulating cell proliferation, Apoptosis, and differentiation. However, challenges related to drug resistance have emerged in the clinical applications of this family. This study addresses resistance mediated by the L858R/T790M/C797S EGFR mutation through a multilevel virtual screening strategy that integrates 3D shape similarity screening, multitask deep learning-based activity prediction, molecular docking, and molecular dynamics simulations. From 18 million drug-like molecules screened, 12 candidates underwent in vitro enzymatic testing, leading to the identification of three novel scaffold inhibitors. Compound L15 demonstrated potent inhibitory activity against the L858R/T790M/C797S mutant EGFR (IC₅₀ = 16.43 nM), exhibiting 5-fold selectivity over wild-type EGFR (IC₅₀ = 80.96 nM). Additionally, it exhibited comparable efficacy against the d746-750/T790M/C797S variant (IC₅₀ = 16.53 nM). Interaction analysis revealed that L15 stabilizes its binding conformation via dominant hydrophobic interactions with LEU718 and LEU792, as revealed by free energy decomposition. This work establishes a systematic multilevel virtual screening strategy to overcome EGFR resistance, providing structural and mechanistic insights for the rational design of fourth-generation inhibitors.

L858R/T790M/C797S mutant EGFR; epidermal growth factor receptor (EGFR); molecular dynamics simulation; multitask deep learning; virtual screening.