Parallel Fast Linear Space Alignment

We introduce a new parallel algorithm for pairwise sequence alignment, called Parallel FastLSA, that finds optimal answers, uses linear space with respect to problem size and number of processors, and achieves good speedups in practice. Computational biologists use sequence alignment algorithms to compare sequences of DNA (or proteins). Since the sequences can be long, space efficiency and turnaround time can be important, especially for interactive use. We have experimented with Parallel FastLSA on an SGI Origin 2400 shared-memory parallel computer using a variety of mammalian DNA sequences from external research groups. The algorithm achieves good speedups on problems of sufficient granularity. We demonstrate speedups of between 3.73 (small problem) and 7.50 (large problem) on 8 processors, and up to 11.28 on 16 processors.