Hidden Markov models for unaligned DNA sequence comparison

Abstract

Comparison of similarity between sequences can provide information for inferring the function of a newly discovered sequence, and understanding the evolutionary relationships among genes, proteins, and entire species. This paper presents a technique for computing the similarity between unaligned DNA sequences. The computation is based on the Kullback-Leibler divergence of hidden Markov models. We used the data sets taken from the threonine operons of Escherichia coli K-12 and Shigella flexneri to test the proposed method. The result obtained agrees with an alignment-based method. We further tested the proposed method with a data ~et of 34 complete mammalian mtDNA genomes. The phylogenetic tree derived from the second experiment shows reasonable evolutionary relationships between these species.