Identification of Key Metabolite Concentrations and Enzyme Saturations Determining the Physiological States of Glucose-Fed E. coli for the Production of 1,4-Butanediol

Metabolic Engineering is fostered by an increasing effort and progress in the collection of large and accurate ‘omics’ datasets. The abundance of available measurements and advances in measurement techniques only alleviate to a certain extent the uncertainty around the physiological states of an organism. In fact, due to intrinsic complexity of metabolic networks, it is not possible to determine their exact intracellular states such as metabolic flux distributions by integrating only the currently available fluxomics and metabolomics data into a model. For these reasons, it is desirable to have a method that further reduces the uncertainty in depicting the actual intracellular states. To tackle this problem, we propose a novel computational approach, based on Monte Carlo sampling and machine learning classification. We start from the integration of known physiological information, in the form of cultivation data, thermodynamics, kinetic rate laws of the constituent reactions and available kinetic parameters. This allows the formulation of kinetic models that describe the possible behaviors, realized by the integrated information. We exhaustively sampled the space of metabolite concentration levels and enzyme saturations. We performed a consistency check on the models of the analyzed physiology to unravel conditions on these quantities that reflect particular physiological states, reducing even more the space of realizable metabolite concentration levels. We approximated this space with well-determined sub-spaces, generated by classification rules on the metabolite concentrations values. Through this approximation, we achieve two important objectives: (i) the computation of the stable physiological volume, as the sum of the volumes of the sub-spaces; (ii) the characterization of all the topological properties of the space, with the most important being the contiguity of the space. The latter helps to discriminate the quantities that determine the attainment of selected production profiles. We applied the novel approach against reported experimental studies of glucose-fed E. coli for the production of 1,4-butanediol. We have found the metabolite concentrations and enzyme saturations, and the relations between these conditions, which are the most decisive in realizing a specific growth and production phase. The proposed approach offers a new way to identify in advance which quantities should be measured to better characterize a particular physiological state, and it can guide experiments for the improvement of biocatalyst performance.