Nonequilibrium Chimeric Switching (NEX) Stabilizes Binding Free Energy Calculations Across Chemical Space

Accurate and efficient prediction of protein–ligand binding free energies remains a barrier to the routine use of physics-based modeling in drug discovery. Although alchemical free energy methods can achieve high accuracy, their applicability is often limited by computational cost and chemical diversity. Here we present Nonequilibrium Chimeric Switching (NEX), a general free energy framework that leverages nonphysical states as endpoints in nonequilibrium transformations. In this work, we investigate one realization of NEX that integrates nonequilibrium statistical mechanics with a dual-topology alchemical transfer formulation for relative binding free energy calculations. By routing bidirectional nonequilibrium switches through a chimeric intermediate, NEX improves phase-space overlap and stabilizes free energy estimation. NEX achieves accurate and reproducible predictions across protein–ligand benchmarks, with most results within 1 kcal mol⁻¹ of experiment. Comparative analysis shows that the chimeric intermediate is critical, reducing work-distribution variance, improving numerical stability, and outperforming direct nonequilibrium switching at equivalent cost. NEX also delivers competitive performance on a challenging scaffold hopping benchmark without manual intervention or system-specific tuning. The chemical flexibility and numerical stability of NEX position it as a scalable engine for hit-to-lead and lead optimization workflows, enabling reliable deployment in future autonomous, agentic drug discovery systems.