Computational analysis of metabolic modules and pathways in the E.coli metabolic network

Construction of genome-scale models of metabolic networks is an important problem in systems biology. Most of the studies of metabolic networks have focused of steady state flows in the native and mutated networks. However, to understand cellular adaptation and development, regulation of metabolism have to be taken into account. The approach of system biology would be to consider metabolism in a broad genomic context. Here we undertake and in-depth exploration of E.coli metabolism by considering the metabolic network together with the network of gene regulation and co-regulation. E.coli has been selected for this study because, (1) it has a very well mapped metabolic network [2, 4], and (2) numerous close organisms have been sequenced allowing accurate prediction of co-regulation from chromosomal location and co-inheritance [6, 3]. We integrated metabolic network with the genomi-cally predicted gene co-regulation. Co-regulation can also be deduced from expression profiling, such data are not available for E.coli in large amounts and may not be as accurate as genomic prediction that are based of dozens of complete bacterial genomes. We first map the network on a graph in which vertices represent reactions and two types of edges: one that connect reactions sharing a metabolite (neighbors on the metabolic network), and the another that connect pairs of reactions that are catalyzed by co-regulated enzymes [6, 8] (co-regulated reactions). Our goals were to explore regulation of known pathways, reveal co-regulated pathways and modules, and understand principles of regulation of metabolism. To achieve these goals we analyzed the metabolic network on three scales. Macroscale analysis focuses on general properties of the network such as the correlation between pathway distance and enzyme co-regulation. Mesoscale analysis fo-cuses on regulation within and between known metabolic pathways, as well as search for clusters of coregulated reactions with short metabolic distance, novel coregulated modules. The key idea is that a metabolic module or pathway has many edges of both types (see above) between its enzymes and hence constitutes a highly-connected subgraph of interrelated and co-regulated metabolic reactions. Finally, microscale analysis focuses on co-regulated motifs of few enzymes that have a particular architecture of chemical reactions. We use novel algorithms to search for highly-connected sub-graphs in biological networks. These techniques combine exact enumeration of cliques, similar to that applied in [7], Super-paramagnetic Clustering — an algorithm developed by Domany and coworkers to cluster objects in a non-metric space of an arbitrary dimension [1], and novel Monte-Carlo …