LNCS Homepage
CD ContentsAuthor IndexSearch

Evolving Better Multiple Sequence Alignments

Luke Sheneman and James A. Foster

Initiative for Bioinformatics and Evolutionary Studies (IBEST), Department of Computer Science, University of Idaho, Moscow, ID 83844-1010, USA
sheneman@cs.uidaho.edu
foster@cs.uidaho.edu

Abstract. Aligning multiple DNA or protein sequences is a fundamental step in the analyses of phylogeny, homology and molecular structure. Heuristic algorithms are applied because optimal multiple sequence alignment is prohibitively expensive. Heuristic alignment algorithms represent a practical trade-off between speed and accuracy, but they can be improved. We present EVALYN (EVolved ALYNments), a novel approach to multiple sequence alignment in which sequences are progressively aligned based on a guide tree optimized by a genetic algorithm. We hypothesize that a genetic algorithm can find better guide trees than traditional, deterministic clustering algorithms. We compare our novel evolutionary approach to CLUSTAL W and find that EVALYN performs consistently and significantly better as measured by a common alignment scoring technique. Additionally, we hypothesize that evolutionary guide tree optimization is inherently efficient and has less time complexity than the commonly-used neighbor-joining algorithm. We present a compelling analysis in support of this scalability hypothesis.

LNCS 3102, p. 449 ff.

Full article in PDF


lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004