Optimization approaches for the in silico discovery of optimal targets for gene over/underexpression.
Metabolic engineering (ME) efforts have been recently boosted by the increase in the number of annotated genomes and by the development of several genome-scale metabolic models for microbes of interest in industrial biotechnology. Based on these efforts, strain optimization methods have been proposed to reach the best set of genetic changes to apply to selected host microbes, in order to create strains that are able to overproduce metabolites of industrial interest. Previous work in strain optimization has been mostly based in finding sets of gene (or reaction) deletions that lead to desired phenotypes in computational simulations. In this work, we focus on enlarging the set of possible genetic changes, considering gene over and underexpression. A gene is considered under (over) expressed if its expression value is constrained to be significantly lower (higher) than the one in the wild-type strain, used as a reference. A method is proposed to propagate relative gene expression values to flux constraints over related reactions, making use of the available transcriptional/translational information. The algorithms chosen for the optimization tasks are metaheuristics such as Evolutionary Algorithms (EA) and Simulated Annealing (SA), based on previous successful work on gene knockout optimization. These methods were modified appropriately to accommodate the novel optimization tasks and were applied to study the optimization of succinic and lactic acid production using Escherichia coli as the host. The results are compared with previous ones obtained in gene knockout optimization, thus showing the usefulness of the approach. The methods proposed in this work were implemented in a novel plug-in for OptFlux, an open-source software framework for ME. Supplementary Material is available at www.liebertonline.com/cmb.