Benchmarking Generative AI and Physics-Based Molecular Simulation for Sampling Conformational Heterogeneity in T4 Lysozyme

Abstract Wild-type T4 lysozyme (T4L) is used as a benchmark to evaluate conformational sampling across generative AI, AI-accelerated molecular simulation (AMS), and physics-based enhanced molecular dynamics (EMD). A four-state model─exposed/open, exposed/closed, buried/open, and buried/closed; is defined using physically meaningful collective variables. Generative AI methods (AF-cluster, MSA subsampling of AlphaFold2, ConforFold, AlphaFlow, ESMFlow, ConfRover, BioEmu) largely sample only the exposed/open state. In contrast, AMS integrates generative ensembles with physics-based molecular simulations, recovering all states and reproducing equilibrium populations similar to EMD and experimental smFRET signatures.