Accelerating Grid Kernels for Virtual Screening on Graphics Processing Units
S aacute nchez-Linares Irene, P eacute rez-S aacute nchez Horacio, García José Manuel · Advances in parallel computing · 2012
The discovery of new drugs can be drastically accelerated with the use of Virtual Screening (VS) methods, ongoing trend in medical research. VS methods need to screen chemical compound databases with millions of ligands but they are completely constrained by the access to computational resources. In order to solve this problem we worked on the adaptation of their main bottlenecks to GPUs using a grid approach with different interpolation methods. In our first studies for medium size proteins, our CUDA implementation runs around 30 times faster than the sequential counterpart.