Accelerating Protein-Protein Docking using a Graphics Processing Unit (GPU)

MICHAEL JENIK; ESTEBAN MOCSKOS; ADRIÁN ROITBERG; ADRIÁN TURJANSKI

Buenos Aires

Simposio; Simposio de Computación de Alto Rendimiento; 2010

SADIO

Abstract. Solving the structure of protein-protein complexes is one ofthe most important tasks in structural biology. Even though there hasbeen great progress in recent years there still a small number of proteincomplexes structures deposited in the Protein Data Bank in comparisonto isolated partners. In this sense, the computational prediction of proteincomplexes starting from the unbound structures, protein-protein Dock-ing algorithms, has emerged as a reasonable alternative. Many dockingprograms employ Fast Fourier Transform (FFT) correlations as an effi-cient search strategy. We describe an implementation of a protein-proteindocking program based on FFT surface complementarity that runs en-tirely on a Graphics Processing Unit (GPU), including grid generation,rotation and scoring. We evaluate its performance, and show that it canbe up to 13 times faster than conventional CPU based implementations.