Cellular metabolism is very complex. Large scale networks that are used for modeling single-cell organism or tissue-specific systems typically comprise of several thousand reactions, each representing a unique biochemical conversion of substrate to product. These in silico models have the potential for predicting how a cell may respond to a perturbation in the form of either a genetic intervention or external stimulus. However, the sheer complexity of these networks remains an impediment for the construction of predictive kinetic ODE models, because the number of system parameters that need to be estimated typically far exceeds the available experimental data and most estimated parameters are not statistically identifiable. Alternatively, graph-based modeling of metabolic networks, where reactions can be denoted by nodes and their interactions described by directed edges, allow one to survey solely the topology of the network and identify structural features that may offer predictable dynamics. Moreover, graph theoretical tools allow for the discovery of modules, or a subset of reactions containing few inputs and outputs, that together function in concert to isolate perturbations from propagating to the rest of the network, a characteristic of metabolic robustness. In this regard, the systematic modularity analysis serves to reduce the complexity of metabolic models and identify modules that both confer robustness and reveal strong coupling among reactions that may not necessarily be intuitive by viewing a two-dimensional cartography of metabolism.;In this thesis, the governing hypothesis is that retroactive, or cyclical, interactions in the form of feedback loops and metabolic cycles engender robustness, and serve as a defining structural feature for the systematic identification of functional modules. As such, a graph-theoretical metric called the Shortest Retroactive Distance (ShReD) is introduced to be used in conjunction with a known network partition algorithm to produce a hierarchical tree of modules, each enriched in cyclical pathways and allosteric feedback loops. Applied to a hepatocyte (liver cell) metabolic network, the ShReD-based partition identifies a `redox' module that couples reactions from apparently distant pathways such as glucose, pyruvate, lipid, and drug metabolism through the shared production and consumption of NADPH, suggesting that cofactors greatly influence the modularity of the network. Recognizing that metabolic networks are not static, a metabolic flux-based edge weighting scheme is proposed to capture the relative engagement between reaction nodes in the graph network. Applying the ShReD-based partition algorithm to weighted adipocyte (fat cell) networks reveals that major physiological changes such as cellular differentiation lead to substantial reorganization in the modularity of the network. In addition, ShReD-based modularity serves as a platform for a targeted motif search within functional modules to discover novel metabolic substrate cycles (a.k.a. futile cycles), which have been recently proposed to be targets for obesity and even cancer. Identifying these substrate cycles requires elementary flux modes (EFM) computation, which would otherwise be infeasible on a large scale network.;Prospectively, modularity analysis of metabolic networks provides theoretical guidance for which reaction rates and metabolite levels may be altered in the face of a perturbation. To experimentally confirm predictions, targeted metabolomics using tandem mass spectrometry (LC/MS-MS) is used to obtain absolute quantification of metabolite concentrations. As an example, an in silico model predicts a set of tryptophan-derived metabolites that can only be exclusively produced by the gut microbiome and may have anti-inflammatory properties. In vivo levels of these indole-backbone metabolite levels are quantified in cecum samples from mice at two different age groups. Statistically significant differences between the two groups suggest that age influences the microbiome composition as well as the metabolites they produce.
展开▼
