Optimization-based Methodologies and Algorithms for Dynamic Metabolic Flux Analysis

Mathematical models for the growth, survival, inactivation and product formation of microbial organisms are becoming increasingly important for the model-based design, optimization and control of bioreactors in (industrial) biotechnology, and for assessment of food safetynbsp;quality throughout the supply chain in predictive microbiology. However, existing models mostly focus on describing these systemsnbsp;a macroscopic point of view. To further enhance the applicabilitynbsp;robustness of these models, integration of mechanistic knowledge is a necessity. The objective of this research is the development of advanced numerical strategies that take advantage of microscopic, mechanistic knowledge in the form of a biochemical reaction network, to infer intracellular knowledge and to build new generation predictive models. The general framework that is used throughout the work is the dynamic metabolic flux analysis model structure, with its descriptive and predictive modes. In the first part, a novel algorithm for offline, dynamic estimation of metabolic fluxes based on B-spline flux parameterizations is developed. This algorithm addresses the disadvantages of existing methodologies. Furthermore, a self-contained strategy for choosing the set of free fluxes in the dynamic metabolic flux analysis model is described. The algorithm is illustrated on two simulated case studies. The second part describes the application of this algorithm in a relevant predictive microbiology case study involving the estimation of metabolic fluxes during a lag phase induced through a sudden shift in temperature. Two bioreactor experiments, performed to gather the necessarynbsp;for the estimation, are described,nbsp;after the definition of the metabolic reaction network, the fluxes are estimated. The evolution of the different metabolic pathways is analyzed and some interesting patterns can be discerned. Finally, in the last part, also an online methodology for flux estimation based on the predictive dynamic metabolic flux analysis model structure is shown. The methodology uses two black-box kinetic flux models and moving horizon estimation to get continuously updated estimates ofnbsp;fluxes and a predictive model, ready for model-based control. The influence of important methodology parameters is assessed on a small-scalenbsp;study, and the performance of the algorithm with a realistic, medium-scale network is determined.status: publishe