Flux measurement selection in metabolic networks

Genome-scale metabolic networks can be reconstructed using a constraint-based modeling approach. The stoichiometry of the network and the physiochemical laws still enable organisms to achieve certain objectives -such as biomass composition- through many various pathways. This means that the system is underdetermined and many alternative solutions exist. A known method used to reduce the number of alternative pathways is Flux Balance Analysis (FBA), which tries to optimize a given biological objective function. FBA does not always find a correct solution and for many networks the biological objective function is simply unknown. This leaves researchers no other choice than to measure certain fluxes. In this article we propose a method that combines a sampling approach with a greedy algorithm for finding a subset of k fluxes that, if measured, are expected to reduce as much as possible the solution space towards the 'true' flux distribution. The parameter k is given by the user. Application of the proposed method to a toy example and two real-life metabolic networks indicate its effectiveness. The method achieves significantly more reduction of the solution space than when k fluxes are selected either at random or by a faster simple heuristic procedure. It can be used for guiding the biologists to perform experimental analysis of metabolic networks.