This (suggestive) example illustrates why a systematic study of combinatorial subsets of these categories can be interesting for understanding the topological basis of essential reactions. A third category of reactions comes from a sampling of random environmental conditions and predicting

steady-state fluxes that optimize biomass production using FBA. The set Inhibitors,research,lifescience,medical of reactions predicted to be active in all conditions has been termed metabolic core (MC) [21]. Remarkably, the MC and the other two topological reaction categories are all fairly accurate predictors of reaction essentiality. Although experimental data from systematic knockout studies is available for E. coli [22,23], these essentiality profiles result from a limited set of environmental conditions. In particular, it has been pointed out recently that essentiality is often medium-dependent [24,25]. While this has been analyzed in [25] for genetic interactions (i.e., the effect of a knockout under the condition of another Inhibitors,research,lifescience,medical knockout), we analyze here the above categories (SA, UPUC

and MC reactions) in light of single-knockout mediumdependent essentiality. An alternative approach of exploring the relationship Inhibitors,research,lifescience,medical between network architecture and function is based on the enumeration of few-node subgraphs. It has been shown that the subgraph composition of functionally related networks tends to be similar [26]. Also, in some cases, dynamical functions can be explained by small few-node subgraphs serving as devices for specific

tasks organized locally in the graph. A potential signature of the functional role of few-node subgraphs is their statistical over- or under-representation (compared to a suitable Inhibitors,research,lifescience,medical Inhibitors,research,lifescience,medical ensemble of random graphs). Such subgraphs are called network motifs. This general concept has been Cisplatin research buy introduced and developed by the Alon group [27,28], particularly for transcriptional regulatory networks [26,29], but not for metabolic networks. For an analysis of a network motif in the context of metabolism see [30] Here we explore the question if a topological understanding of reaction essentiality can be established by integrating the in silico determined knock-out data with the three reaction categories and all Thymidine kinase combinatorial three-node subgraphs. We start by introducing the relative essentiality of a reaction defined on the basis of a large number of combinatorial minimal media simulations. For each medium, the essentiality of all active reactions is tested in silico. In Section 2.1 the relative essentialities will be used as a basis of the three essentiality classes: always essential (essential), essential only in some growth media (conditional lethal), and never essential (non-essential). Section 2.2 is devoted to an initial analysis of the three categories of reactions (UPUC, SA and MC).