Genome-scale metabolic networks are often computationally explored to charac­terize functional relationships between their reactions. For instance, identification of correlated reaction sets (cosets), i. e., reactions that are always “on” or “off” con­currently (Papin et al. 2004) can define functional relationships between reactions that are not necessarily in the same pathway or obvious. The significance of these reactions is evident from the observation that mutations in correlated reactions may lead to a manifestation of the same aberrant (or disease) phenotype (Jamshidi and Palsson 2006). Because, the solution space of genome-scale networks can be

enormous, uniform sampling of the space is often carried out using the Monte Carlo method to identify the cosets and the overall shape and size of the steady state flux space (Becker et al. 2007). This sampling method, which is implemented in COBRA, identifies a set of randomly distributed solutions to serve as a proxy for the entire space. In Monte Carlo sampling, points are picked randomly from the space and the fraction inside the defined constraints is counted. This sampling method allows a uniform exploration of the metabolic network space while reducing the computational power demand required for the analysis.