NEMAT: An Automated Non-Equilibrium Free-Energy Framework for Predicting Ligand Affinity in Membrane Proteins

Abstract

Quantifying the strength of small-molecule binding to proteins is essential for understanding biological function and for advancing drug discovery. Computational free-energy methods aim to predict these interactions accurately, yet membrane proteins remain particularly difficult due to the added thermo-dynamic effects of the lipid bilayer. Alchemical free-energy calculations provide state-of-the-art accuracy by transforming one ligand into another, and recent non-equilibrium approaches (NEQ-FEP) offer improved efficiency and parallelization. However, no existing workflow automates NEQ-FEP for membrane proteins and enables decomposition of the observed binding free energy into membrane-partitioning and protein-specific components. NEMAT is an open-source pipeline that performs automated non-equilibrium alchemical transformations in water, membranes, and membrane-embedded protein environments. Here we show that NEMAT reproduces experimental binding-energy trends for P2Y1 ligands with accuracy comparable to established equilibrium methods. These findings demonstrate that NEQ-FEP can be applied reliably to membrane systems when supported by a consistent workflow. In contrast to previous approaches, NEMAT provides systematic control over transition length, sampling, and replica averaging, enabling stable work-distribution overlap and reproducible free-energy estimates. In a broader context, NEMAT offers a practical route to mechanistically interpretable affinity predictions for membrane-embedded targets. Its ability to dissect membrane and protein contributions advances our understanding of lig- and selectivity at lipid-facing pockets and supports the routine application of non-equilibrium free-energy methods in drug discovery.