Transcriptional regulation in Escherichia coli: A systems biology approach

High-throughput technologies are yielding large data sets that require network-based data analysis to reconcile heterogeneous data types, find inconsistencies, and systematically generate hypotheses. To begin this process in Escherichia coli, a genome-scale model of its metabolic and transcriptional regulatory networks was reconstructed. Although genome-scale constraints-based models of Escherichia coli metabolism have already been constructed and used to successfully interpret and predict cellular behavior under a range of conditions, such models do not account for regulation of gene transcription and thus cannot accurately predict some organism functions. I therefore expanded the constraint-based approach to modeling metabolic systems to incorporate transcriptional regulatory events in terms of time-dependent constraints on the metabolic network. The effects of regulatory constraints on metabolic behavior were studied using extreme pathway and flux-balance analysis. I demonstrate that the imposition of environmental conditions and regulatory mechanisms sharply reduces the number of active extreme pathways, and that incorporation of transcriptional regulatory events in flux-balance analysis enables interpretation, analysis and prediction of the effects of transcriptional regulation on cellular metabolism at the systemic level. The genome-scale metabolic and regulatory model accounts for 1,010 genes and was used to computationally predict growth phenotypes of 110 knockout strains under 125 growth conditions (13.750 cases). The computations were consistent with experimental measurement in 10.828 (79%) cases, and resolution of discrepancies between prediction and observation led to identification of 18 areas where the metabolic or regulatory networks are incompletely characterized. To begin further characterization of the regulatory network, I also mRNA expression profiled wild-type and 6 knockout strains under aerobic and anaerobic conditions. Altered expression of 151 genes represented in the model were detected and 22 of these changes were due entirely to regulation that had been previously described. Model-driven analysis of the remaining cases led to the formulation of 110 new regulatory rules that represent testable hypotheses. Overall, I find that a systems biology approach that combines genome-scale experimentation and computation can systematically generate hypotheses from disparate data sources.