SWAMP+: Enhanced Smith-Waterman Search for Parallel Models

Abstract

More sensitive than heuristic methods for searching biological databases, the Smith-Waterman algorithm is widely used, but it has a high quadratic running time. This work presents a faster approach and implementation for Smith-Waterman that extends the traditional results to return multiple, BLAST-like sub-alignments. The extended Smith-Waterman using Associative Massive Parallelism (SWAMP+) is introduced for three different parallel architectures: Associative Computing (ASC), the Clear Speed coprocessor, and the Convey Computer FPGA coprocessor. We show that parallel versions of Smith-Waterman can be successfully modified to produce multiple BLAST-like sub-alignments while maintaining the original Smith-Waterman sensitivity. This approach combines parallelism and the novel extension to produce multiple sub-alignments for pair wise comparisons. The two parallel SWAMP+ implementations for the ASC model and the Clear Speed CSX-620 use a wave front approach. Both perform a full trace back in the parallel memory and return multiple subsequence alignment results. Results show a linear speedup for the 96 processing elements (PEs) on a single Clear Speed chip. The third approach is a SWAMP+ adaptation that uses the non-associative Convey FPGA coprocessor. This allows for an initial high-speed, high-throughput Smith-Waterman alignment on the hybrid system optimized for large databases. The additional pair wise alignments are run to produce the additional SWAMP+ sub-alignments. The overall results across the three systems are parallel implementations of an extended Smith-Waterman that maintain a speedup and provide a deeper exploration of the query sequences not previously available.