Rapid design of system-wide metabolic network modifications using iterative linear programming

Computationally-aided metabolic engineering is an important, complementary strategy to combinatorial strain design for enhanced biochemical production by microbes. Bilevel optimization problems have been formulated for optimal strain design via reaction removal, activation, and inhibition. Deterministic global optimization of the resulting mixed integer linear programs (MILPs) requires extensive computational effort, especially for genome-scale models of metabolism. Improving the computational efficiency of such algorithms is an ongoing challenge. Here, we present Enhancing Metabolism with Iterative Linear Optimization (EMILiO)–a novel bilevel optimization-based algorithm that includes all possible flux modifications and is solved with remarkable computational efficiency via iterative linear programming. The resulting solution is recursively pruned to generate alternate, parsimonious strain designs with maximal biochemical production rates. We demonstrate our algorithm for succinate production using the iAF1260 genome-scale model of Escherichia coli metabolism. Our algorithm identifies aerobic succinate-producing strains with increased glyoxylate shunt activity, which is consistent with experiments in the literature. We also identified novel strain design strategies that may have implications for the control of industrial bioreactors to maximize succinate production.