Faster Sampling in Molecular Dynamics Simulations with TIP3P-F Water
A significant challenge for using MD to sample configurations and determine equilibrium properties of biomolecular systems, is the computation required to sufficiently sample the timescales necessary to observe the relevant thermodynamic minima. In order to sample biomolecular events within reasonable timeframes, computational chemists often employ enhanced sampling or reduce the precision of the model to coarse-grained definitions. Even so, our ability to study events such as protein folding or ligand binding is limited by what we can capture in a few microseconds of simulation. Frustratingly, much of the computation is spent on the water molecules that solvate our systems, and it is the high-frequency motions in the system that limit the MD timestep to 2 fs. Many of us have experienced the catastrophic crashes caused by instable systems when longer timesteps are used.
In this paper of the month, Jimenez et. al. describe and validate TIP3P-F, a “fast” water model. The model uses the concepts of mass repartitioning and mass scaling to modify the masses of the atoms in the water molecules, which decreases the viscosity of the water and stabilises the time integration. The paper demonstrates that using the TIP3P-F model in a neat water system; or to solvate peptides, proteins, nucleic acids and lipid systems, results in negligible differences in the thermodynamic properties. Importantly, the autocorrelation of these properties was observed to reduce by up to 2-fold, indicating a faster convergence in sampling. For studies that require effective sampling of equilibrium properties, as opposed to the observation of actual dynamics, TIP3P-F is another significant contribution in reducing the energy cost and climate impact of MD simulations.
Jimenez et. al. Faster Sampling in Molecular Dynamics Simulations with TIP3P-F Water.
J. Chem. Theory Comput. 2024, 20, 11068-11081