After a year-long effort, we finally developed an efficient implementation of the Gaussian-based volume and surface area model for GPUs. The work, performed in collaboration with Peter Eastman Vijay Pande at Stanford University, is described in a recent publication to appear in the Journal of Computational Chemistry:
Baofeng Zhang, Denise Kilburg, Peter Eastman, Vijay S. Pande, Emilio Gallicchio. Efficient Gaussian Density Formulation of Volume and Surface Areas of Macromolecules on Graphical Processing Units. J. Comp. Chem. (2017). pdf of submitted manuscript
Volume and surface area of macromolecules are employed in medicinal chemistry to measure structural similarity and complementarity of compounds. They are also the basis of many implicit models of non-polar solvation, which is our primary interest.
With this algorithm we were able to achieve a 50- to 100-fold speed-up on GPU's relative to our best CPU implementation. more ->
The GaussVol code is freely available on github as a plugin of the OpenMM molecular mechanics package.
News & Events >