A simple theory for finding related sequences by adding probabilities of alternative alignments

Abstract

SummaryThe main way of analyzing genetic sequences is by finding sequence regions that are related to each other. There are many methods to do that, usually based on this idea: find alignments that are unlikely to occur by chance between unrelated sequences. This approach is suboptimal, and we do not fully understand how to tell if an alignment is likely to occur by chance. A more powerful way is to integrate a similarity measure over alternative ways of aligning the regions. This is rarely done: we lack a clear and simple theory suitable for wide adoption.Here is described a simplest-possible change to standard sequence alignment, which sums probabilities of alternative alignments. It is easy to calculate the probability of such a similarity occurring by chance. This approach is better than standard alignment at finding distant relationships, at least in a few tests. It fits easily as a component in alignment software, and it generalizes to different kinds of alignment (e.g. DNA-versus-protein with frameshifts). Thus, it can widely contribute to finding subtle relationships between sequences.