Accelerating Biological Sequence Alignment Algorithm on GPU with CUDA

In this paper, we have used Compute Unified Device Architecture (CUDA) GPU to accelerate pair wise sequence alignment using the Smith-Waterman (SW) algorithm. Smith-Waterman(SW) is by far the best algorithm for its accuracy in similarity scoring. But the executing time of this algorithm is too long in sequence alignment. So we describe a multi-threaded parallel design and implementation of the Smith-Waterman (SW) on CUDA to reduce execution time. And according the architecure of CUDA, we have divided the computation of a whole pair wise sequence alignment scoring matrix into multiple sub-matrices, using 32 threads to process on submatrice, more over we optimized memory distribution scheme, and used reduction to find the maximum element of the alignment scoring matrix. We experiment the algorimthm on GeForce 9600 GT, connet to Windows xp 64-bit system. The results show this mplementation achieves more better performance than the other parallel implementation on the Graphics Processing Unit.