Basing protein-ligand optimizations on molecular mechanics force fields yields high  accuracy at the cost of speed.  We address this dilemma by circumventing unnecessary calculations and by harvesting the computational power of graphics processor units (GPUs). This novel computational architecture is  well suited for parallelizing and massively speeding  up force-field-based minimization procedures. This endeavor encompasses fundamental adaptions of formerly CPU-based methods in order to take the GPU-specific programming paradigms into account. In our case, this results in force-field kernels uniquely targeted at the efficient optimization of protein-ligand complexes in the context of a mostly GPU-based procedure. On docking poses generated by TrixX[1] with complexes originating from the Astex diverse set[2], our procedure performs at least 100 times faster than widely used CPU-based  optimization tools. Compared to these, it furthermore minimizes docked ligand structures without loss of quality: on the data set mentioned above it reduces the average RMSD value of the ligands from 2.35� down to 1.95�.[1] Schlosser, J.; Rarey, M. Beyond the Virtual Screening Paradigm: Structure-Based Searching for New Lead Compounds. Journal of Chemical Information and Modeling 2009, 49,800�809.[2] Hartshorn, M.; Verdonk, M.; Chessari, G.; Brewerton, S.; Mooij, W.; Mortenson, P.;Murray, C. Diverse, high-quality test set for the validation of protein-ligand docking performance. Journal of Medicinal Chemistry 2007, 50, 726�741.

Structure-based lead optimization: Force-field-based minimizations in the blink of an eye

244th National Meeting (2012)

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