Methods For Predicting Gene Knockout Targets Based On Metabolic Network Analysis

Construction of artificial cell factory which can produce specific compound of interest needs wild strain to be genetically engineered frequently. In recent years, with the reconstruction of many genome-scale metabolic networks, a number of strain optimization methods based on metabolic network analysis for predicting genetic manipulation targets that lead to overproduction of compounds of interest have been proposed. These approaches make use of constraints of stoichiometry and reaction reversibility in genome-scale models of metabolism and adopt different mathematical algorithms to predict modification targets by coupling biomass formation with chemical production. However, they cannot predict manipulation targets for most of compounds of interest due to the fact that the production of these metabolites can't be coupled with cellular growth. Moreover, their running time grows exponentially with the number of manipulations resulting in finding an optimal solution is difficult.The currently computational tools for flux variability analysis are too inaccurate to qualitatively classify reactions, so we develop a new tool-QCFVA(http://www.ibiodesign.net/newfva/) which overcomes the problem of calculation accuracy of the exiting methods. Based on metabolic network analysis, we propose a new strain optimization method-FVAKnock, which can predict gene knockout targets by considering the trade-off between growth and production. In this new framework, we aim to identify genes as knockout targets after deleting which biomass growth is reduced while the optimal target product synthesis pathway isn't affected by making use of QCFVA to qualitatively classify reactions in metabolic network. FVAKnock is applied to a large-scale E. coli metabolic network iJO1366 for gene deletion target prediction. Dozens of genes are identified as deletion targets for the production of poly-?-hydroxybutyrate(PHB), succinate and riboflavin respectively. Previous methods such as OptKnock could not successfully find deletion targets for these metabolites such as PHB and riboflavin because their production cannot be coupled with cellular growth. After testing the FVAKnock for different target products, we demonstrate that the new strain optimization method is faster and more efficient than the previous methods. Moreover, the prediction results are biologically significant.In addition, this paper integrated boolean transcriptional regulatory network in E. coli integrated network model iMC1010v2 into genome-scale E. coli metabolic network iJO1366 by making use of r FBA. Then, we develop a new method-rFVA, which can be used to determine the minimum and maximum flux values that the reactions in integrated network can carry. Finally, based on integrated network and rFVA, we propose a new strain optimization method-rFVAKnock, which can predict gene knockout targets that lead to the optimal production of compounds of interest.