Comparative analysis of module-based versus direct methods for reverse-engineering transcriptional regulatory networks.
BackgroundThe transcriptional regulatory network of Escherichia coli has been extensively studied and is considereda standard reference network for evaluating network inference algorithms. The module network approach is awell-known probabilistic network inference strategy which aims to identify condition-specific regulatory programsfor modules of coexpressed genes from gene expression data.
ResultsWe have applied an improved, ensemble-based module network algorithm to a benchmark E. coli expressioncompendium and compared it to the CLR algorithm which aims to identify individual transcription regulatoryinteractions. The complementarity of these opposite approaches is shown by the fact that the union of bothnetworks performs better than each network separately, which in turn each perform better than the intersection.We interpret the module network in terms of the hierarchical topological structure of the E. coli reference networkand show that it predominantly infers local regulators in the bottom layers of the hierarchy. Many of thesehigh-scoring regulators constitute neighbor regulators, consistent with the viewpoint that network evolution in thebottom layers is mainly driven by horizontal gene transfer.
ConclusionsThe ensemble-based module network approach infers high-quality modules and condition-specificregulatory programs in bacteria which mainly model the bottom layers of the hierarchical transcriptional regulatorynetwork. These bottom layers probably contain most of the uncharacterized regulators and the module networkapproach is therefore well suited to predict putative targets of these regulators for which no other data butgene expression data is available. Comparison of the module network and CLR methods suggests that futurereverse-engineering improvements will lie in the integration of different types of models.
Michoel, T., De Smet, R., Joshi, A., Van de Peer, Y., Marchal, K. (2009) Comparative analysis of module-based versus direct methods for reverse-engineering transcriptional regulatory networks. BMC Syst. Biol. 3:49.
VIB / UGent
Bioinformatics & Evolutionary Genomics
+32 (0) 9 33 13807 (phone)
+32 (0) 9 33 13809 (fax)