Probabilistic pairwise sequence alignment

Lawren Smithline

13 November 2003

Abstract:

We describe an new algorithm for visualizing an alignment of biological sequences according to a probabilistic model of evolution. The resulting data array is readily interpreted by the human eye and amenable to digital image techniques.

We present examples using mRNA sequences from mouse and rat: three cytochromes and two zinc finger proteins. The underlying evolutionary model is derived from one proposed by Thorne, Kishino, and Felsenstein and improved by Hein and others. The demonstration implementation aligns two sequences using time and memory quadratic in the mean sequence length.

The algorithm is extensible, after Lunter, Miklós, Song and Hein to multiple sequences. We mention a basic method to reduce time and memory demands.