DISCOVERY OF A NOVEL PHENYLAMINE-PURINE CHEMOTYPE AS A POTENTIAL EGFR INHIBITOR USING AN INTEGRATED AI-DRIVEN DRUG DESIGN PIPELINE

Abstract

The discovery of novel Epidermal Growth Factor Receptor (EGFR) inhibitors remains critical for overcoming resistance in targeted oncology. We report a fully integrated, artificial intelligence-driven drug discovery pipeline combining generative modeling, structure-based virtual screening, bioactivity prediction, and molecular dynamics. Using the generative deep-learning model DrugGen, we autonomously designed over 8,000 novel compounds, filtering candidates through hierarchical molecular docking and the BOLTZ-2 deep-learning activity predictor. This methodology successfully identified a structurally unique phenylamine-purine scaffold (PPH) that diverges from traditional quinazoline-based inhibitors. PPH demonstrated a highly favorable docking score (−10.811) compared to the reference drug Erlotinib (−9.261), establishing critical stabilizing hydrogen bonds with Asp831, Met769, and Lys721 within the ATP-binding pocket. Subsequent 100 ns molecular dynamics simulations confirmed the long-term dynamic stability of the PPH-EGFR complex. Furthermore, in silico ADME analysis validated the scaffold's favorable drug-like properties, establishing the AI-generated phenylamine-purine core as a highly promising, novel chemotype for next-generation EGFR inhibition.