Electrostatic free energy calculations using the generalized solvent boundary potential method

Abstract

Free energy perturbation (FEP) calculations using all-atom molecular dynamics simulations with a large number of explicit solvent molecules are a powerful approach to study ligand–macromolecule association processes at the atomic level. One strategy to carry out FEP calculations efficiently and reduce computational time is to consider the explicit dynamics of only a small number of atoms in a localized region around the ligand. Such an approximation is motivated by the observation that the factors governing binding specificity are dominated by interactions in the vicinity of the ligand. However, a straightforward truncation of the system may yield inaccurate results as the influence exerted by the remote regions of the macromolecule and the surrounding solvent through long-range electrostatic effects may be significant. To obtain meaningful results, it is important to incorporate the influence of the remote regions of the ligand–macromolecule complex implicitly using some effective potential. The generalized solvent boundary potential (GSBP) that was developed recently [W. Im, S. Bernèche, and B. Roux, J. Chem. Phys. 114, 2924 (2001)] is an efficient computational method to represent the long-range electrostatic interactions arising from remote (outer) regions in simulations of a localized (inner) region with a small number of explicit atoms. In the present work, FEP calculations combined with GSBP are used to illustrate the importance of these long-range electrostatic factors in estimation of the charging free energy of an aspartate ligand bound to the aspartyl-tRNA synthetase. Calculations with explicit spherical simulation inner regions of different radii are used to test the accuracy of the GSBP method and also illustrate the importance of explicit protein and solvent dynamics in the free energy estimation. The influence of the represented outer region is tested using separate simulations in which the reaction field and/or the protein static field are excluded. Both components are shown to be essential to obtain quantitatively meaningful results. The ability of implicitly treating the influence of protein fluctuations in the outer region using a protein dielectric constant is examined. It is shown that accurate charging free energy calculations can be performed for this system with a spherical region of 15 to 20 Å radius, which roughly corresponds to 1500–3500 moving atoms. The results indicate that GSBP in combination with FEP calculations is a precise and efficient approach to include long-range electrostatic effects in the study of ligand binding to large macromolecules.