Modeling Biological Networks

• IV.6.iv Subproject 4 Modeling the Regulation and Function of Metabolism
  • IV.6.iv.a Datasources
  • IV.6.iv.b Methods
  • IV.6.iv.c Methodology
    • IV.6.iv.c.1 Cluster analysis
    • IV.6.iv.c.2 Uncovering dynamical constraints using Singular Value Decomposition
  • IV.6.iv.d Research Plan
    • IV.6.iv.d.1 Model construction
    • IV.6.iv.d.2 Model validation
      • IV.6.iv.d.2.i Phenotypic validation
      • IV.6.iv.d.2.ii Molecular validation
      • IV.6.iv.d.2.iii Experimental validation
• [ Other Subprojects ]

< Previous | Page 24 of 35 | Next >

The skeleton of our model is the stochiometric matrix Sij, which encodes the topology of the metabolic network. S will be the starting point for analyzing and locating the pathway structure as well as for flux balance analysis. Information on the identity of the enzymes that catalyze the appropriate reactions will complement this matrix. The metabolic reactions and enzymes together form a bipartite graph containing two qualitatively different categories of nodes, with nodes in one category linking only to nodes in the other category (Strogatz, 2001). One projection of this bipartite graph recovers the metabolic network, another the enzyme network, in which the metabolites that participate in catalyzed reactions connect enzymes.

If Subproject 1 confirms the modularity of metabolism, we will assign each reaction and metabolite further variables indicating their modules, allowing us to track the extent of metabolic perturbations. In addition, a separate enzymatic correlation matrix, E, will carry information about the regulatory correlations between the enzymes that Subproject 1 uncovers from the microarray data. This will quantify the extent to which changes in one segment of the network diffuse to other components along the regulatory pathways, which we treat semi-quantitatively as a separate layer. We assign the components of this matrix (which are the enzymes in the bipartite graph) additional variables based on the clustering the expression data uncovers. An important issue is how to couple the regulatory network E matrix to the bipartite metabolic topology network. The regulatory network affects the flux balance by modulating the available number of enzymes that can catalyze reactions. Thus a simple approach is to include regulation-imposed limits on the fluxes, wherever we observe a strong suppression or activation correlation in coregulation. We will investigate other sorts of coupling, including nonlinear relationships. In some cases, such as evaluating changes in flux balances under gene knockout, a strong hypothesis about the nature of the coupling between the two layers is essential. However, for evaluating the extent of damage or lethality, the nature of the coupling may not be crucial.

As the utility of the model depends on its ability to predict experiments, validation will be an extensive part of our program:

• Phenotypic validation, aimed at determining whether genes are essential.
• Molecular validation, predicting gene expression for cells under extreme conditions and following the removal of an enzyme's coding region.
• Validation based on emerging experimental information.

Validation will feed back to the model, which will incorporate the additional information the validation generates.

To test the reliability of our models we will predict the essentiality of individual genes in E. coli. The inhomogeneous network connectivity of the metabolic and protein networks affects the robustness of metabolic networks. For instance, scale-free networks are insensitive if randomly chosen nodes malfunction (Albert et al., 2000). However, the price of this robustness is extremely diminished attack survivability. Deleting the most connected nodes rapidly degrades the ability of the remaining nodes to communicate. For a cell, scale-free metabolism is robust against most random errors, but elimination of a highly connected component is lethal.

In a metabolic network, several mechanisms can alter the connectivity. Since enzymes catalyze biochemical reactions, eliminating a particular enzyme (e.g., through gene ablation or pharmacological inhibition) will affect the corresponding link(s) between nodes. Second, as many substrates or products of metabolism require transport for entry (or exit), elimination of their corresponding transporter will directly affect the corresponding substrate (node). Finally, removing a gene can perturb the genetic network, which can affect other parts of the metabolic network. To address the fault tolerance of a metabolic network under these separate attack modes, we propose an integrated experimental and theoretical program.