Systems biology uses mathematical and computational methods to describe and explore complex biological networks. An important recent trend in systems biology has been the development and application of constraint-based modeling.1 Constraint-based modeling provides 3 significant advantages compared with traditional mathematical approaches for the study of large and complex biochemical systems. First, large models (up to thousands of reactions) can be accommodated. Thus, the entire metabolic network of a mitochondrion, for example, can be modeled. Second, precise descriptions of the behavior of each enzyme in the system (ie, rate laws) are not required. Finally, detailed information regarding the activity of a single protein (eg, whether an enzyme is allosterically modified or not) is not necessary. Thus, unlike traditional kinetic models, constraint-based models do not rely on, nor do they require, detailed knowledge of an enzyme's phosphorylation status, for example, nor the abundance of substrates and products.
Clinical Perspective on p 415Constraint-based modeling is able to confer these advantages because the underlying models and assumptions are simple. The basic unit for constraint-based modeling is a network model, similar to the London Underground map. In the case of a biochemical network, this is constructed using (1) the known presence or absence of reactions based on genomic, proteomic, or biochemical data and (2) the known (speciesspecific) stoichiometry of all the chemical reactions included in the network. To this basic model are added a series of constraints (from which the method derives its name), including Background-Any reduction in myocardial oxygen delivery relative to its demands can impair cardiac contractile performance. Understanding the mitochondrial metabolic response to hypoxia is key to understanding ischemia tolerance in the myocardium. We used a novel combination of 2 genome-scale methods to study key processes underlying human myocardial hypoxia tolerance. In particular, we hypothesized that computational modeling and evolution would identify similar genes as critical to human myocardial hypoxia tolerance. Methods and Results-We analyzed a reconstruction of the cardiac mitochondrial metabolic network using constraintbased methods, under conditions of simulated hypoxia. We used flux balance analysis, random sampling, and principal component analysis to explore feasible steady-state solutions. Hypoxia blunted maximal ATP (−17%) and heme (−75%) synthesis and shrank the feasible solution space. Tricarboxylic acid and urea cycle fluxes were also reduced in hypoxia, but phospholipid synthesis was increased. Using mathematical optimization methods, we identified reactions that would be critical to hypoxia tolerance in the human heart. We used data regarding single-nucleotide polymorphism frequency and distribution in the genomes of Tibetans (whose ancestors have resided in persistent high-altitude hypoxia for several millennia). Six reactions were identified by both methods as being criti...