ABSTRACT
Stoichiometric models, such as those used in ux balance analysis (FBA), have also emerged as powerful analysis tools that couple observed extracellular phenomena (uptake/production rates, growth rate, product and biomass yields, etc.) with the intracellular carbon ux and energy distribution. Constraints based stoichiometric models do away with kinetics in favor of a pseudo-steady-state picture of metabolism. FBA and stoichiometric models have been employed to calculate genomic-scale snapshots of several organisms as well as portraits of key subnetworks such as central carbon metabolism. One of the rst examples of what would evolve into FBA was the analysis of butyric-acid bacteria by Papoutsakis [36-38]. Later, Varma, Palsson, and coworkers employed a stoichiometric model of E. coli W3110 to study oxygen limitation and by-product secretion [39,40]. Vallino and Stephanopoulos employed FBA to explore Corynebacterium glutamicum during lysine overproduction [41,42], while Sauer et al., characterized the metabolic capabilities of riboavin producing B. subtilis [43]. Pramanik and Keasling explored the impact of time varying biomass composition and E. coli metabolism [44,45] while Maranas and coworkers explored the performance limits of E. coli subject to gene additions or deletions [46], the coupling of metabolic uxes in large-scale networks [47], the generation of optimal gene deletion strategies [48], the production of lactic acid in E. coli [49] and the computational identication of reaction activation/inhibition or elimination candidates in metabolic networks [50]. Edwards, Schilling, Palsson, and coworkers extended FBA to genomic-scale metabolic reconstructions of Helicobacter pylori 26,695 (389 reactions) [51], E. coli MG1655 (740 reactions) [52,53], E. coli K-12 (931 reactions) [54], S. cerevisiae (1173 reactions) [55] and most recently to the human metabolic map with a genome scale reconstruction consisting of 3,311 metabolic and transport reactions and 2,766 metabolites [56]. An attractive feature of constraints based models is the relative ease of computation (solving a linear program or determining a matrix inverse) and the ability to directly incorporate process information, for example on-line CO2, O2, or cellmass measurements into the constraints (see Savinell and Palsson for discussion of optimal measurement selection [57] or Becker et al., for FBA soware [58]). In addition to physiological measurements, 13C-NMR/GC-MS labeling techniques have been employed by many groups to add additional constraints to the ux calculation [59-74]. Sauer et al., (and others) have pushed 13C enhanced metabolic ux estimation beyond serial experiments into the realm of parallel high-throughput data generation; see Ref. [75].
