How are metabolic network models reconstructed? A metabolic network consists of metabolites, biochemical reactions, and the relevant genomic evidence for the described enzymatic reactions, or gene-protein-reactions (GPRs) associations. The structural framework of a genome-scale metabolic model begins with compilation of relevant gene annotations and ends with refinement of the reconstructed meta­bolic network. This reconstruction process passes through four blocks (Fig. 10.3) (Orth et al. 2010; Thiele and Palsson 2010) as described in the sections below.

10.4.1 Draft Reconstruction

In this step (Fig. 10.3a), stoichiometric reactions that can describe cellular metab­olism using various sources of information are compiled. These data may be col­lected from different knowledgebases, including BiGG (http://bigg. ucsd. edu) (Schellenberger et al. 2010), KEGG (http://www. genome. jp/kegg/), MetaCyc (http://www. metacyc. org), and peer reviewed literature. Through genomic and bioinformatics approaches, the functional annotations of open reading frames (ORFs) provide genomic evidence for the presence of specific biochemical reac­tions, associating genes, or multiple genes with specific reactions in the network, for generating the gene-protein-reaction associations. Gene products are associated with specific reactions through assigned enzyme commission (EC) numbers. This process may include sequence-based searches of the ORFs against well-curated databases such as UniProt (http://www. uniprot. org) (Apweiler et al. 2004) or through profile-based scans (e. g., InterPro, http://www. ebi. ac. uk/interpro/) (Jones et al. 2014) to assign enzymatic function and EC numbers to the ORFs.

Model Refinement

Fig. 10.3 Metabolic model reconstruction and refinement. a Information from one or more knowledgebases is extracted to define reactions and pathways to reconstruct a draft model; b the draft network model is transformed into a stoichiometric matrix that maps the metabolites and the associated reactions; c the obtained mathematical representation of metabolism is constrained with key flux parameters and can be optimized for an objective function. The obtained optimal solution is then validated by experimental data; differences between the two are reconciled by refining the initial model through filling gaps, adding and removing metabolites, obtaining additional experimental evidence

With enzymatic functions assigned, metabolic reactions can be defined which in turn allows reconstruction of a draft metabolic network. The draft network also accounts for metabolites that contribute to biochemical reactions inside the cell.