MCALIGN2: Faster, accurate global pairwise alignment of non-coding DNA sequences based on explicit models of indel evolution

Background: Non-coding DNA sequences comprise a very large proportion of the total genomic content of mammals, most other vertebrates, many invertebrates, and most plants. Unraveling the functional significance of non-coding DNA depends on how well we are able to align non-coding DNA sequences. However, the alignment of non-coding DNA sequences is more difficult than aligning protein-coding sequences. Results: Here we present an improved pair-hidden-Markov-Model (pair HMM) based method for performing global pairwise alignment of non-coding DNA sequences. The method uses an explicit model of indel length frequency distribution which can be specified, and allows any time reversible model of nucleotide substitution. The method uses a deterministic global optimiser to find the alignment with the highest posterior probability. We test MCALIGN2 in simulations, and compare it to a previous Monte Carlo based method (MCALIGN), to the pair HMM method of Knudsen and Miyamoto, and to a heuristic method (AVID) that performed very well in a previous simulation study. We show that the pair HMM methods have excellent performance for all combinations of parameter values we have considered. MCALIGN2 is up to ten times faster than MCALIGN. MCALIGN2 is more accurate in resolving indels given an accurate explicit model than heuristic methods, but is computationally slower. Conclusion: MCALIGN2 produces better quality alignments by explicitly using biological knowledge about the indel length distribution and time reversible models of nucleotide substitution. As a result, it can outperform other available sequence alignment methods for the cases we have considered to align non-coding DNA sequences.