An efficient way of finding good indel seeds for local homology search

Abstract

Designing good or optimal seeds is a key factor for local homology search in bioinformatics. Continuous seeds have existed for nearly 20 years used by BLAST series programs. Recently, spaced seeds, which were introduced by PattenHunter program, were shown to be more sensitive and faster than continuous seeds under the same similarity level. However, there are 2 main disadvantages for space seeds: (i) It assumes that only matches and mismatches occur within seed alignments, but not insertions and deletions (indels); (ii) calculating optimal spaced seeds is an NP-hard problem. Introduction for indel seeds solved the first problem, but the second is getting much harder because of its higher exponential level. In this paper, we introduce an efficient way of designing good (even optimal) indel seeds under “indel overlap complexity” model, and it can be calculated in polynomial time. We calculate indel seeds from weight of 11 to 15. The result shows that indel seeds have higher sensitivities than spaced ones and our algorithm finds good indel seeds very quickly.