Chapter 5 - Using Genetic Algorithms for Pairwise and Multiple Sequence Alignments

Abstract

One of the most commonly used techniques in bioinformatics is the simultaneous alignment of many nucleic acid or amino acid sequences. Given a set of homologous sequences, multiple alignments are used to help predict the secondary or tertiary structure of new sequences, to help demonstrate homology between new sequences and existing families, to help find diagnostic patterns for families, to suggest primers for the polymerase chain reaction (PCR), and as an essential prelude to phylogenetic reconstruction. These alignments may be turned into profiles or Hidden Markov Models (HMM) that can be used to scour databases for distantly related members of the family. Multiple alignment techniques can be divided into two categories: global and local techniques. When making a global alignment, the algorithm attempts to align sequences chosen by the user over their entire length. Local alignment algorithms automatically discard portions of sequences that do not share any homology with the rest of the set. Most multiple alignment methods are global, leaving it to the user to decide which portions of the sequences are to be incorporated. To aid that decision, researchers often use local pairwise alignment programs, such as BLAST or a straightforward implementation of the Smith and Waterman algorithm. This chapter focuses on global alignment methods with a special emphasis on the alignment of protein and RNA sequences. It introduces the concept of evolutionary computation for optimization of multiple sequence alignments.