SW#db: GPU-accelerated exact sequence similarity database search

Abstract

The deluge of next-generation sequencing (NGS) data and expanding database poses higher requirements for protein similarity search. State-of-the-art tools such as BLAST are not fast enough to cope with these requirements. Because of that it is necessary to create new algorithms that will be faster while keeping similar sensitivity levels. The majority of protein similarity search methods are based on a seed-and-extend approach which uses standard dynamic programming algorithms in the extend phase. In this paper we present a SW#db tool and library for exact similarity search. Although its running times, as standalone tool, are comparable to running times of BLAST it is primarily designed for the extend phase where there are reduced number of candidates in the database. It uses both GPU and CPU parallelization and when we measured multiple queries on Swiss-prot and Uniref90 databases SW#db was 4 time faster than SSEARCH, 6-10 times faster than CUDASW++ and more than 20 times faster than SSW.