Applying GPUs for Smith-Waterman Sequence Alignment Acceleration

Abstract

The Smith-Waterman algorithm is a common local sequence alignment method which gives a high accuracy. However, it needs a high capacity of computation and a large amount of storage memory, so implementations based on common computing systems are impractical. Here, we present our implementation of the Smith-Waterman algorithm on a cluster including graphics cards (GPU cluster) -swGPUCluster. The algorithm implementation is tested on a cluster of two nodes: a node is equipped with two dual graphics cards NVIDIA GeForce GTX 295, the other node includes a dual graphics cards NVIDIA GeForce 295 and a Tesla C1060 card. Depending on the length of query sequences, the swGPUCluster performance increases from 37.33 GCUPS to 46.71 GCUPS. This result demonstrates the great computing power of GPUs and their high applicability in the bioinformatics field.