Edward and Palsson developed the first genome-scale metabolic reconstruction in 1999 for Haemophilus influenza (Edwards and Palsson 1999) and described its emerging systems properties through flux balance analysis (FBA). In FBA, fluxes of metabolites through biochemical reactions are constrained by four parameters: mass conservation, thermodynamics (reaction reversibility), steady state assumption for internal metabolite concentrations, and nutrient availability. All of these constraints define the boundary conditions required to solve systems of linear equations in which a biologically motivated objective function (e. g., biomass production) is optimized.

The solution of an FBA problem is the optimal distribution of fluxes through the metabolic network, comprising a metabolic phenotype or functional state of the network (Orth et al. 2010). Assumptions and boundary flux parameters can reduce the size of the feasible solution space. An objective function, which is a set of reactions that the cell is assumed to be optimizing for a given mode of growth, can be mathematically (through linear optimization) maximized (or minimized) to derive flux values that support optimal solution(s) for the metabolic state under investigation. The solution is of course sensitive to growth and boundary param­eters for a given model. As an example, iRC1080 reconstruction of C. reinhardtii metabolic network (Chang et al. 2011) was optimized for biomass production under two different conditions of growth, with light and acetate, and dark with acetate. As the alga grows in the dark, it relies solely on acetate as a source of energy and carbon, while photosynthesis provides both under light growth. Consequently, major flux redistributions can be expected system-wide and are observed between simulated growths under these two conditions (Fig. 10.4).

Metabolic models can make use of high-throughput data such as gene expression data (including mRNA and protein expression data), and 13C flux data by directly imposing additional constraints on the metabolic model based on the values obtained from wet-bench experiments. For example, if obtained experimental data show that glycolytic enzymes are highly active under certain conditions, flux can be pointed through glycolysis by constraining the relevant fluxes in silico, thus driving flux through the activated reactions and allowing estimation of changes in global flux distributions (Shlomi et al. 2005). Phenotypic data, such as Biolog data (Bochner 2003, 2009), which describe cellular metabolic profiles, can also be used to validate in silico phenotypes (Oberhardt et al. 2008). Notebaart and his col­leagues used Saccharomyces cerevisiae as a model organism for this type of analysis. In their work, they carried out a comparison of in silico metabolic fluxes versus microarray gene expression data in E. coli and S. cerevisiae. Their results revealed that metabolic genes whose fluxes are directionally coupled generally

show similar expression patterns, share transcriptional regulators, and reside in the same operon (Notebaart et al. 2008).

Flux variability analysis (FVA), which is a variant of FBA, determines the max­imum and minimum values of all the fluxes that will satisfy the constraints and allow for the same optimal objective value (Gudmundsson and Thiele 2010). For example, FVA can be applied to predict the range of possible by-product production rates under maximal biomass production, which can be linked to gene expression data. Varia­tions of FVA can also be used to determine blocked or nonessential reactions.