Exploring a coarse-grained distributive strategy for finite-difference Poisson-Boltzmann calculations

Abstract

We have implemented and evaluated a coarse-grained distributive method for finite-difference Poisson-Boltzmann (FDPB) calculations of large biomolecular systems. This method is based on the electrostatic focusing principle of decomposing a large fine-grid FDPB calculation into multiple independent FDPB calculations, each of which focuses on only a small and a specific portion (block) of the large fine grid. We first analyzed the impact of the focusing approximation upon the accuracy of the numerical reaction field energies and found that a reasonable relative accuracy of 10-3 can be achieved when the buffering space is set to be 16 grid points and the block dimension is set to be at least (1/6)3 of the fine-grid dimension, as in the one-block focusing method. The impact upon efficiency of the use of buffering space to maintain enough accuracy was also studied. It was found that an "optimal" multi-block dimension exists for a given computer hardware setup, and this dimension is more or less independent of the solute geometries. A parallel version of thedistributive focusing method was also implemented. Given the proper settings, the distributive method was able to achieve respectable parallel efficiency with tested biomolecular systems on a loosely connected computer cluster. © 2010 The Author(s).