Nucleic Acids Research Advance Access published online on July 7, 2008
Nucleic Acids Research, doi:10.1093/nar/gkn407
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Computational Biology |
MOPAT: a graph-based method to predict recurrent cis-regulatory modules from known motifs
1Division of Biostatistics, 2School of Informatics and 3Center for Computational Biology and Bioinformatics, School of Medicine, Indiana University, 410 West 10th Street, Indianapolis, IN 46202, USA
*To whom correspondence should be addressed. Tel: +1 317 278 7273; Fax: +1 317 278 9217; Email: shawnli{at}iupui.edu
Received January 26, 2008. Revised June 1, 2008. Accepted June 10, 2008.
The identification of cis-regulatory modules (CRMs) can greatly advance our understanding of eukaryotic regulatory mechanism. Current methods to predict CRMs from known motifs either depend on multiple alignments or can only deal with a small number of known motifs provided by users. These methods are problematic when binding sites are not well aligned in multiple alignments or when the number of input known motifs is large. We thus developed a new CRM identification method MOPAT (motif pair tree), which identifies CRMs through the identification of motif modules, groups of motifs co-ccurring in multiple CRMs. It can identify ‘orthologous’ CRMs without multiple alignments. It can also find CRMs given a large number of known motifs. We have applied this method to mouse developmental genes, and have evaluated the predicted CRMs and motif modules by microarray expression data and known interacting motif pairs. We show that the expression profiles of the genes containing CRMs of the same motif module correlate significantly better than those of a random set of genes do. We also show that the known interacting motif pairs are significantly included in our predictions. Compared with several current methods, our method shows better performance in identifying meaningful CRMs.