Speeding up pairwise sequence alignments: A scoring scheme reweighting based approach

A general technique based on scoring scheme reweighting is proposed that can be used to speed up dynamic programming algorithms for a variety of pairwise sequence alignment problems. For the standard sequence alignment problem with an arbitrary gap penalty function, we show that a reweighted scoring scheme can be obtained by an efficient preprocessing step that computes a set of upper bounds on the score of the optimal alignment between pairs of suffixes of the sequences. A series of experiments on synthetic sequences and biological sequences indicate that our algorithm offers significant and robust speedup over the standard cubic-time dynamic programming algorithm. For sequences of length up to 2000 used in our experiments, the speedup factor ranges from 4 to more than 50. With a strong upper bound, a sub-cubic behavior in running time is also observed for all the tested situations.