Quantitative analysis of Poisson-Boltzmann implicit solvent in molecular dynamics.

A critical issue in the development of implicit solvent models is their quality in realistic simulations of non-trivial systems. In a previous study, we quantitatively compared the reaction field energies of static structures calculated with the Poisson-Boltzmann implicit solvent and the TIP3P explicit solvent and found an overall agreement, though a discrepancy was also observed in the electrostatic potentials of mean force for salt-bridging and hydrogen-bonding dimers (see J. Phys. Chem. B, 2006, 110, 18680). In this study, we are interested in how the implicit solvent performs in molecular dynamics simulations. To guarantee sampling convergence in simulated observables in the explicit solvent, we explored to use a high-temperature constant-volume simulation setting at 450 K but with the water density at 300 K. The relevance of the artificial simulation setting to room-temperature simulations of biomolecules was first investigated by systematic comparisons of the polar and nonpolar solvation free energies of 23 amino acid analogues at 300 K and 450 K, respectively. Assisted by the artificial simulation setting, we found the simulated secondary structure populations agree very well between the implicit and explicit solvents for tested dipeptides and peptides. In addition, the agreement in the populations of hydrophobic contacts is reasonable. However, our analysis also shows that the populations of the salt bridges are too low in the implicit solvent. The low salt-bridge population perhaps results from a combination of the atomic-centered modified van der Waals surface and the small solvent probe radius optimized to best reproduce the polar potential of mean force profiles. In addition, the lower accuracy of the electrostatic forces and the lack of water-bridged minima in the implicit solvents may also contribute to the instability of the salt bridge populations.