(72f) Large Scale Distributed Computing Calculations of Binding Free Energies Using Folding@Home and GPU Simulations of Proteins Using OpenMM | AIChE

(72f) Large Scale Distributed Computing Calculations of Binding Free Energies Using Folding@Home and GPU Simulations of Proteins Using OpenMM

Authors 

Shirts, M. R. - Presenter, University of Virginia


We first present our efforts to use distributed computing to calculate high precision free energies of binding using distributed computing. We discuss our simulations of small molecule binding to FKBP, where we obtain binding free energies less than 1 kcal/mol RMS from experiment for molecules of typical drug size, with sampling validated by the fact that relative free energies with the same system differ from absolute free energies of binding by less than 0.3 kcal/mol.  We also present our results with the SAMPL3 blind test set, using distributed computing generalized ensemble simulations to identify binders to trypsin out of a set of 540 small molecules, and to compute the free energies of binding for known binders.  In both cases, 10,000's of volunteer computers on the Folding@Home distributed computing system were used.  We also present our efforts to study local protein unfolding using implicit solvent simulations on GPUs using the open source OpenMM API, which allows simulations up to 70x faster than standard packages on CPU's.