StapleBridge: Chemistry-Aware Schrödinger Control for Hydrocarbon-Stapled Peptide Lead Optimization

Hydrocarbon stapling can improve peptide conformational stability, proteolytic resistance, and cellular uptake, but computational peptide design has largely focused on de novo generation rather than optimizing an existing functional lead. Lead-specific stapling instead requires jointly selecting anchor positions, staple chemistry, and minimal residue substitutions under explicit geometric constraints, despite scarce paired linear-to-stapled supervision. We introduce StapleBridge, a chemistry-aware discrete Schrödinger-control framework for hydrocarbon-stapled peptide lead optimization. For each lead, StapleBridge constructs a finite space of executable stapling interventions. A canonical plan-product mapping makes the soft-terminal control target exactly computable, providing direct lead-specific supervision without paired stapling examples and enabling amortized intervention selection on unseen leads. On a held-out benchmark, StapleBridge consistently outperforms matched constrained baselines in predicted permeability improvement and intervention-selection quality, while substantially reducing Plan Regret. Post-hoc 3D relaxation further suggests improved backbone and helicity preservation. These results support structured intervention control as an effective formulation for chemically constrained peptide lead optimization.