Multiple sequence alignment

A multiple sequence alignment extends pairwise alignment to three or more sequences at once. Each sequence becomes a row, gaps are inserted to represent insertions and deletions, and each column ideally holds residues descended from the same position in a common ancestor. The result shows at a glance which positions are conserved across all the sequences and which vary.

Finding the best possible alignment of many sequences is computationally hard, because the work grows explosively with each sequence added, so programs use shortcuts. Progressive methods, such as Clustal Omega, first build a guide tree from pairwise comparisons and then join sequences and groups in order, starting with the most similar. Iterative tools such as MUSCLE and MAFFT refine an initial alignment repeatedly, consistency-based tools such as T-Coffee check pairwise alignments against one another, and methods based on hidden Markov models score alignments with probabilities. Results are often viewed and edited in programs such as Jalview.

Multiple alignments are the starting point for building phylogenetic trees, spotting conserved residues likely to matter for a protein's structure or function, defining protein families and domains, and designing PCR primers in conserved regions. Mistakes in the alignment carry through to every later step, so results are worth checking by eye.