Text S 1 ( Analog regulation of metabolic demand )

The strength of graph theory is that it can represent a complex system in a unified formal language of nodes and links. Examples of graph-theoretical analyses of metabolic systems are [1] and [2]. Essentially, the pattern of zero and non-zero entries of the stoichiometric matrix defines a graph representation of the metabolic system. The level of information conveyed by looking at metabolism from a graph-theoretical perspective is still subject to constant scrutiny [3]. A suitable approach for analyzing the correspondence between expression changes and metabolism and thus quantifying metabolic coherence is the application of the tools developed for the effective TRNs [4] to a gene-centric representation of metabolism, where the nodes are metabolic genes and a link is drawn between two genes, if the associated reactions share a common metabolite [5]. This representation is the gene-centric variant of one of the standard projections of a bi-partite graph representation of metabolism, where both, metabolites and reactions serve as node sets (see, e.g., 6, for a discussion of these and other representations). Additional forms of representation include metabolite-, reaction-, enzymeor gene-centric views. The metabolite-centric view represents the interconversion possibilities of the different substrates, whereas other views concentrate more on the processes (reactions), protein (enzymes) and genomic (genes) levels respectively. We chose the gene-centric view for our analysis as it allowed us a direct comparison of expression patterns with metabolic pathways.