Optimization of Gaussian surface calculations and extension to solvent-accessible surface areas

We explored the use of several breadth-first and depth-first algorithms for the computation of Gaussian atomic and molecular surface areas. Our result s for whole-molecule van der Waals surface areas (vdWSAs) were 10 times mor e accurate in relative error, relative to actual hard-sphere areas, than th ose reported by earlier workers. We were also able to extend the method to the computation of solvent-accessible surface areas (SASAs). This was made possible by an appropriate combination of algorithms, parameters, and prepr ocessing steps. For united-atom 30app, a 2366-atom protein, we obtained an average absolute atomic error of 1.16 Angstrom(2). With respect to the hard -sphere atomic SASA results in 7 s of CPU time on an R10000/194 MHz process or. Speed and accuracy were both optimized for SASA by the use of neighbor- list reduction (NLR), buried-atom elimination (BAE), and a depth-first sear ch of the tree of atomic intersections. Accuracy was further optimized by t he application of atom type specific parameters to the raw Gaussian results . (C) 1999 John Wiley & Sons, Inc. J Comput Chem 20: 688-703, 1999.