BPS2026 – A new enhanced sampling technique for accelerating amino acid membrane permeation simulations

Efficient conformational sampling on complex free-energy landscapes without being hindered by hidden barriers often requires a combination of parallel tempering techniques and reaction coordinate-based extended-ensemble methods. Here, we developed a new approach, gREST-ABMD, which integrates generalized replica exchange with solute tempering (gREST) and adaptively biased molecular dynamics (ABMD). We implemented this method in the GENESIS molecular dynamics simulation package and applied it to membrane permeation simulations of several amino acids. Furthermore, we compared two different solute regions in gREST—one including only the amino acid itself and the other including both the amino acid and the membrane head groups. Compared with the widely used umbrella sampling method, gREST-ABMD required only about one-tenth as many replicas, thereby substantially reducing the computational cost for a single free-energy calculation. Despite this reduction, the free-energy landscapes obtained by gREST-ABMD were comparable in accuracy to those obtained by umbrella sampling. For amino acids with complex aromatic side chains, calculations that defined the solute region as the amino acid alone did not converge even within 1 μs of simulation, whereas including the membrane head groups in the solute region enabled convergence of the free-energy landscape within a few hundred ns. These results demonstrate that the newly developed gREST-ABMD method provides an efficient approach to obtain converged free-energy landscapes for peptide membrane permeation with reduced computational resources, provided that an appropriate solute region is defined.