Motif-based Hidden Markov Models for Multiple Sequence Alignment

Motif-based Hidden Markov Models for Multiple Sequence Alignment (poster)

Protein families are well characterized by a collection of motifs (Sonnhammer & Kahn 1994), sometimes referred to as the ``common core'' (Chothia & Lesk 1986).
These motifs can have structural and functional significance, and they may frequently be operated upon as units by diverse evolutionary mechanisms.
The quality of a multiple alignment depends upon how accurately it identifies an ordered series of motifs (McClure et al. 1994, 1995).
Hidden Markov models (HMMs) provide a theoretically sound modeling paradigm for collections of motifs for which efficient algorithms exist.
Meta-MEME (Grundy et al. 1996, 1997) is a software toolkit that builds left-to-right, motif-based HMMs that focus upon the common core.
Meta-MEME has been shown to detect remote homologies using smaller training sets than are required by standard HMMs.
In an analysis of four protein families, Meta-MEME alignments are shown to be of higher quality than those produced by standard HMMs.
In a previous analysis of nine other multiple alignment methods (McClure et al. 1994), only one method yields significantly higher quality alignments than Meta-MEME.