Comparing algorithms for large-scale sequence analysis

Abstract

The first step in homology analysis is usually the comparison of sequences by similarity search. The explosive growth of genomic databases makes it increasingly important to develop more rapid approaches to the comparison of large sequence databases while using the most sensitive methods available. This paper explores the consequences of this trade-off, comparing the results produced by BLAST and Smith-Waterman on genoinic- scale sequence searches. Stich comparisons are now possible thanks to the development of novel distributed computing platforms. This study uses the Parabon Frontier/sup TM/ Internet computing platform, which enables the effective use of the vast supply of idle computer cycles on the Internet for high-performance computing. We have ported both Smith-Waterman and BLAST to the Frontier platform, enabling the efficient use of these algorithms on large sequence databases. In addition, we present a novel visualization tool along with quantitative metrics for comparing the results of alternative sequence alignment algorithms. Our results compare the sensitivity of Smith-Waterman and BLAST for identifying homologies on proteome databases.