Multi-objective optimization of genome-scale metabolic models: the case of ethanol production

Abstract

Ethanol is among the largest fermentation product used worldwide, accounting for more than 90 % of all biofuel produced in the last decade. However current production methods of ethanol are unable to meet the requirements of increasing global demand, because of low yields on glucose sources. In this work, we present an in silico multi-objective optimization and analyses of eight genome-scale metabolic networks for the overproduction of ethanol within the engineered cell. We introduce MOME (multi-objective metabolic engineering) algorithm, that models both gene knockouts and enzymes up and down regulation using the Redirector framework. In a multi-step approach, MOME tackles the multi-objective optimization of biomass and ethanol production in the engineered strain; and performs genetic design and clustering analyses on the optimization results. We find in silico E. coli Pareto optimal strains with a knockout cost of 14 characterized by an ethanol production up to 19.74mmolgDW-1h-1 (+832.88% with respect to wild-type) and biomass production of 0.02h-1 (-98.06%). The analyses on E. coli highlighted a single knockout strategy producing 16.49mmolgDW-1h-1 (+679.29%) ethanol, with biomass equals to 0.23h-1 (-77.45%). We also discuss results obtained by applying MOME to metabolic models of: (i) S. aureus; (ii) S. enterica; (iii) Y. pestis; (iv) S. cerevisiae; (v) C. reinhardtii; (vi) Y. lipolytica. We finally present a set of simulations in which constrains over essential genes and minimum allowable biomass were included. A bound over the maximum allowable biomass was also added, along with other settings representing rich media compositions. In the same conditions the maximum improvement in ethanol production is +195.24%.