Recently the quick development of GPU-accelerated computing has seen significant impacts on scientific, analytics, engineering, consumer, and enterprise applications. In the area of molecular simulations for drug design problems, GPU-accelerated high performance computing has proven to be able to provide the computing power that once only was achievable through expensive supercomputers, at only a small fraction of cost. Nevertheless, accurate free energy methods on GPU have been implemented in only very few simulation packages.<br/>Here we report the implementation of the thermodynamics integration method (TI) on the pmemd module of AMBER 16 package on GPU. The pmemd with GPU-enhanced TI (pmemdGT) utilizes the object-oriented programming principles so that the newly add-on codes are inherited from, not direct extension of, the original pmemd gpu codes. As a result, the TI GPU part is separated from the original code hence much easier to maintain, test, as well as to add new scientific algorithm.<br/>Benchmarks with selected protein-ligand systems demonstrate that, when compared to single CPU core, pmemdGT is capable of delivering more than 100X speed-up and essentially the same binding affinities with errors of 0.1 kcal/mole.

pmemdGT: An efficient and accurate GPU-accelerated thermodynamics integration simulation package

253rd National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.