A reproducible molecular dynamics benchmark for ADC epitope triage in EGFR and HER2

Antibody–drug conjugate programs often commit to an epitope before peptide–level experimental evidence is available. I present a reproducible molecular–dynamics benchmark for pre–HDX epitope triage in EGFR and HER2, evaluated on three therapeutic antibody–antigen pairs (HER2–Trastuzumab, EGFR–Cetuximab, and EGFR–Panitumumab). For each system, paired simulations of antigen alone and antibody–bound antigen are summarized into ranked antigen regions suitable for designing an initial HDX–MS peptide panel. Performance is quantified with rank–based enrichment metrics appropriate for imbalanced labels, together with uncertainty estimates where panel size is small. For systems with published peptide–level HDX–MS labels (HER2–Trastuzumab and EGFR–Cetuximab), ranked regions are compared to reported protection patterns. For EGFR–Panitumumab, where peptide–level HDX–MS labels are not used here, evaluation is performed against proxy labels derived from structural contacts and mutational mapping of the EGFR domain III footprint. This work is framed as a reproducible benchmarking and triage layer rather than a predictor of absolute deuterium uptake or clinical efficacy. Success is defined as enrichment of known or proxy–positive regions near the top of the ranked list under a locked evaluation protocol.