Large scale protein sequence alignment using FPGA reprogrammable logic devices

Abstract

In this paper we show how to significantly accelerate Smith-Waterman protein sequence alignment algorithm using reprogrammable logic devices - FPGAs (Field Programmable Gate Array). Due to perfect sensitivity, the Smith-Waterman algorithm is important in a field of computational biology but computational complexity makes it impractical for large database searches when running on general purpose computers. Current approach allows for aminoacid sequence alignment with full substitution matrix which leads to more complex formula than used in DNA alignment and is much more memory demanding. We propose different parellization scheme than commonly used systolic arrays, leading to full utilization of PUs (Processing Units), regardless of sequence length. FPGA based implementation of Smith-Waterman algorithm can accelerate sequence alignment on a Pentium desktop computer by two orders of magnitude comparing to standard OSEARCH program from FASTA package. © Springer-Verlag Berlin Heidelberg 2004.