(441b) A Novel Optimization Strategy for Microbial Strain Design
Metabolic engineering has emerged as an important field aimed to improve cellular production of valuable biochemicals and biofuels. Conventional metabolic engineering approaches focus on metabolic branch points, and eliminate undesired reactions in competing branches and enhance flux through desired reactions using genetic modifications. However, these metabolic network modifications not only affect local metabolic pathways, but also have global effects on metabolic behavior due to changes in carbon, energy, and electron flows. Correspondingly, such conventional approaches may miss modifications in distant pathways that can potentially improve the cellular function of interest.
Computational approaches have been successful in identifying genetic modification strategies for overproduction of various biochemicals using genome-scale metabolic models. These approaches often rely on the combinatorial optimization algorithms that search for k perturbations from a total of n reactions. The complexity of such algorithms increases exponentially with k where the base of exponential is n. As a result, the problem becomes intractable when k gets larger even with small n. Alternative approaches have been developed where metabolic models are pre-processed to reduce search space and different search algorithms such as genetic algorithms or local heuristics are used to handle the intractability. However, these approaches sometimes converge to local optima and fail to identify the global optimum due to the nature of search algorithms used. Furthermore, the search space for removal of reactions can be significantly different from the search space for removal of genes.
Here, we present a fast and effective approach to identify genetic modification strategies using a mixed-integer programming formulation and solution methods. We have recently developed a new approach (OptORF) which searches for genetic perturbation strategies in genotype space where transcriptional regulation is accounted for. Using the OptORF formulation, we demonstrate how our method finds globally optimal strategies efficiently while other approaches fail. We have applied penalty to perturbations and tightened the bounds for dual variables, and solved successive problems with increasing number of allowed perturbations. We have found that our method can identify global solutions significantly faster than other approaches, and also identify optimal strategies for some difficult cases which were previously not found. Finally, we apply the method to overproduction of biofuels in Escherichia coli.