Multiple Alignment using A* Algorithm and Parallel Iterative Algorithm

Yasushi Totoki(1), Yutaka Akiyama(2), Tamotsu Noguchi(2),
Kentaro Onizuka(2), Minoru Saito(2), Makoto Ando(2)

Multiple sequence alignment is an important and widely used technique for sequence similarity analysis in molecular biology. Since the problem requires enormous calculation time and memory space, several approximation methods have been proposed. However the quality of alignments by the previous methods are limited. As a new strategy, we employed an iterative scheme with best-first search, and parallelized its search step using PVM library. Furthermore we implemented the A* pruning algorithm instead of dynamic programming, to drastically reduce the search space. As a result, our new parallel system enables biologically accurate multiple sequence alignment performed within reasonable time.


(1) Information and Mathematical Science Laboratory, Inc. (2) Real World Computing Partnership