Tools for Analyzing Network Structure and Performance

In general, individual metabolic fluxes are the net result of the coordinated activity of the whole network and so rational manipulation of these fluxes requires tools that can analyze the network as a system rather than focusing on individual steps. The available modeling approaches can be classified on the basis of their underlying assumptions (Wiechert, 2002), and the resulting hierarchy matches the usefulness of the models for metabolic engineering.

The simplest models are the structural network models that are based on the metabolites and reaction steps that make up the network (Wiechert, 2002). Models of this kind are useful for exploring the architecture of the network, but they are of rather limited use in a physiological context because they lack quantitative information about the metabolites and reaction steps. This deficiency is remedied in stoichiometric models by assuming constant fluxes and intracellular pool sizes. Stoichiometric models provide the basis for determining intracellular fluxes (Bonarius et al., 1997), as well as permitting the identification of fundamental network properties such as elementary flux modes and extreme pathways (Klamt and Stelling, 2003). Stoichiometric modeling can also be applied at the level of the individual carbon atoms in metabolites, and this leads to a more general method of determining intracellular fluxes based on the steady-state analysis of the redistribution of 13Clabels (Kruger et al., 2003; Wiechert, 2001; Wiechert et al., 2001).

Models that provide an explanation of the empirically derived flux distribution can be obtained by incorporating a kinetic description of each reaction step into a stoichiometric model (Wiechert, 2002). These mechanistic (kinetic) models require detailed information about the in vivo kinetic properties of the enzymes in the network, and this is a major obstacle in developing useful models. However, kinetic modeling is now well developed in yeasts (Teusink et al., 2000) and red blood cells (Mulquiney and Kuchel, 2003). Accurate mechanistic models are expected to have predictive value in the context of metabolic engineering, and they can also be used to investigate the distribution of control within the conceptual framework of metabolic control analysis (Fell, 1997). Mechanistic models can be used to analyze both steady-state and transient fluxes and in the longer term it may also be possible to allowfor fluctuations in enzyme level by incorporating the regulatory networks for gene expression (Wiechert, 2002). It is clear from this survey that the analysis of the properties of metabolic networks can be approached using a variety of model-based strategies. Some of these approaches aim to make deductions about the performance of the network from an analysis of the constraints imposed by its structure and stoichiometry alone, whereas others are heavily dependent on direct measurements of metabolic fluxes and the kinetic properties of the enzymes that define the network. The aim here is to describe four of these methods in more detail and to comment on their utility as predictive tools for plant metabolic engineering.