Generative Replica Exchange: A Flow-Based Framework for Accelerating Replica-Exchange Simulations

Replica exchange (REX) is one of the most widely used enhanced sampling methodologies. However, its efficiency is often limited by the requirement for a large number of intermediate temperature replicas. Here, we present Generative Replica Exchange (GREX), an enhanced sampling approach that integrates deep generative models into the REX framework to eliminate the need for this temperature ladder. Drawing inspiration from reservoir replica exchange (res-REX), GREX utilizes trained normalizing flows to generate high-temperature configurations on demand and map them directly to the target distribution. To achieve this, we implement the potential energy as a constraint in GREX, eliminating the need for training data at the target temperature. This approach reduces production simulations to a single replica run at the target temperature while maintaining a Metropolis-filtered acceptance step for generated proposals. We validated GREX on benchmark systems of increasing complexity, highlighting its superior efficiency and practical applicability for molecular simulations. To further demonstrate its applicability to larger biomolecular systems, we applied GREX to bovine pancreatic trypsin inhibitor (BPTI), a 58-residue protein with slow conformational dynamics for which conventional REX and related methods face computational challenges.