ALIGNMENT OF THREE SEQUENCES IN QUADRATIC SPACE Xiaoqiu Michigan Abstract W e present a dynamic programming algorithm for computing an optimal global al/gument o f three sequences in quadratic space. The algorithm has been implemented as a portable C program. The program makes it possible to align simultaneously any three protein sequences on workstations with tens of megabytes o f main memory. This rigorous program is used as a benchmark to evaluate the performance of a heuristic multiple sequence alignment program on protein sequences from the Protein Identification Resource protein sequence database. Experimental results indicate that the rigorous program may perform significenfly better than the heuristic program on three sequences of weak similarity. W e also describe algorithms for computing local alignments among three sequences in quadratic space. K e y w o r d s : sequence comparison, dynamic programming, protein sequences Introduction Existing programs for aligning multiple sequences are based on a dynamic programming algorithm for aligning two sequences [2-3, 5, 7-8, 14-16]. One way to improve these programs is to use simultaneous alignment of three sequences as a basic step. However, existing algorithms for simultaneous alignment-of three sequences require O (N 3) time and space,
/lp/association-for-computing-machinery/alignment-of-three-sequences-in-quadratic-space-iOBvyqYqEO