Iterative Multiscale Molecular Dynamics: Accelerating Conformational Sampling of Biomolecular Systems by Iterating All-Atom and Coarse-Grained Molecular Dynamics Simulations

All-atom molecular dynamics (AA MD) simulations remain the gold standard for simulating biomolecular dynamics at an atomistic level for decades. However, they often cannot attain the time scales required to sufficiently sample the many biological processes of interest, especially those that involve large biomolecular systems with high energy barriers. Here, we combine the strengths of all-atom (AA) and coarse-grained (CG) MD simulations to develop the iterative multiscale MD (iMMD) simulation workflow in OpenMM by iterating between AA and CG MD simulations to enhance the sampling of biomolecular conformations. The AA-CG-AA iterations are repeated over multiple cycles, facilitating the accelerated sampling of biomolecular conformations at a CG level, while the finer atomistic interactions are refined with AA simulators. We demonstrate the enhanced sampling ability of iMMD over multiple AA-CG-AA iterations on four representative systems, including two globular proteins (fast-folding variant of Trpcage and Z-matrix protein of Mammarenavirus lassaense (LASV)) and two membrane proteins (Torpedo nicotinic acetylcholine receptor (nAChR) and REGN7663 Fab binding to the CXCR4 receptors in a heterogeneous cholesterol/phosphatidylcholine membrane lipid bilayer). In particular, iMMD captures the folding of the fast-folding Trpcage and Z-protein of LASV starting from their extended conformations as well as the binding of the REGN7663 Fab binding to the CXCR4s and dimerization of the CXCR4 receptors within the heterogenous membrane, within few AA-CG-AA iterations. The development of iMMD should allow for unprecedented simulations of biomolecules at much longer biological timescales to explore the biological processes of interest.