Implicit solvent models are one of the standard tools in computational biophysics. While Poisson–Boltzmann methods offer highly accurate results within this framework, generalized Born models have been used due to their higher computational efficiency in many (bio)molecular simulations, where computational power is a limiting factor. In recent years, there have been remarkable advances to reduce some deficiencies in the generalized Born models. On the other hand, these advances come at an increased computational cost that contrasts the reasons for choosing generalized Born models over Poisson–Boltzmann methods. To address this performance issue, we present a new algorithm for Born radii computation, one performance critical part in the evaluation of generalized Born models, which is based on a Barnes–Hut tree code scheme. We show that an implementation of this algorithm provides accurate Born radii and polar solvation free energies in comparison to Poisson–Boltzmann computations, while delivering up to an order of magnitude better performance over existing, similarly accurate methods. The C++ implementation of this algorithm will be available at http://www.int.kit.edu/nanosim/.
Powerborn: A Barnes-Hut tree implementation for accurate and efficient Born radii computation
M. Brieg and W. Wenzel
Journal of Chemical Theory and Computation 9(3), 1489-1498