Large-scale DFT calculations in implicit solventA case study on the T4 lysozyme L99A/M102Q protein

Recently, variants of implicit solvation models for first principles electronic structure calculations based on a direct solution of the nonhomogeneous Poisson equation in real space have been developed. These implicit solvation models are very elegant from a physical point of view as the solute cavity is defined directly via isosurfaces of the electronic density, and the molecular charge is polarized self-consistently by the reaction field of the dielectric continuum which surrounds the solute. Nevertheless, the implementation of these models is technically complex and requires great care. A certain level of care is required from users of such models as a number of numerical parameters need to be given appropriate values to obtain the most accurate and physically relevant results. Here, we describe in what parts of the solvent model each of these numerical parameters is involved and present a detailed study of how they can affect the calculation, using the solvation model which has been implemented in the ONETEP program for linear-scaling density functional theory (DFT) calculations. As ONETEP is capable of DFT calculations with thousands of atoms, we focus our investigation of the numerical parameters with a case study on proteinligand complexes of the entire 2602-atom T4 Lysozyme L99/M102Q protein. We examine effects on solvation energies and binding energies, which are critical quantities for computational drug optimization and other types of biomolecular simulations. We propose optimal choices of these parameters suitable for routine production calculations. (c) 2012 Wiley Periodicals, Inc.