Some genes are thought to be organized into functional modules, which are switched on in certain stages of development or in response to an environmental condition.
Such coregulated groups of genes will share a similar expression pattern over any range of tissues. Using publicly available microarray expression data, potential modules were identified by agglomerative hierarchical clustering. Each module was evaluated for enrichments of GO(Gene Ontology) classifications and regulatory sequence motifs from the TRANSFAC database. To further examine the conditions
under which a certain module is active, a novel method for visualizing relationships between clusters using linear combinations of expression profiles was developed.
The added expression profiles of modules representing genes specific to e.g. different muscle types will resemble the profile of a more general module, active in several types of muscle. Performing this analysis on a global scale resulted in an interactive map, providing a clear view of how clusters relate to each other in terms of linear combinations. The map was integrated with the other results and made available through a web interface. The expression data used covered a wide range of tissues in mouse and human, providing an excellent opportunity for mammalian module discovery. A large proportion of the inferred modules contained genes which were clearly functionally related. Some represented cellular processes such as protein synthesis or respiration, while others were ”differentiation type” clusters, containing e.g. smooth muscle specific genes. Statistical testing revealed overrepresentation of known cis-regulatory motifs in several modules. |