Research Projects

Hyaluronic acid (HA) is an important biopolymer with applications in the biomedical and cosmetic industries. A particular application of HA highly depends on its molecular weight, often called its grade. Producing the desired grade of HA in industries in bioreactors from glucose is challenging as it is difficult to maintain the glucose (substrate) and acetate concentrations at desired values. It has been known that feedback control can improve the performance of bioprocesses. In this project, we propose to develop model predictive control (MPC) strategies to maximize HA production of the desired grade using Near-infrared spectroscopy-based measurements and constraint-based models. Particularly, this project will investigate finding optimal control strategies that combine information of process-level information (extracellular) obtained by NIR spectroscopy and intracellular metabolism obtained by genome-scale metabolic models for higher HA titers of a desired grade in the presence of process disturbances and parametric changes. The goal of the project is achieved through three objectives. In the first objective, the dynamic flux balance analysis (DFBA) will be applied to understand the effect of process-level changes on intracellular metabolism using the developed genome-scale metabolic models and the kinetic models of extracellular metabolites for the production of HA of a desired grade. Since DFBA models cannot be used directly for an online application due to the large size of reactions and computational cost, it is proposed to develop reduced constraint-based metabolic (RCM) models of HA production under different process conditions and will be validated with experimental data. In the second objective, an MPC problem will be formulated to maximize HA productivity of a desired grade for the fixed batch time subject to the RCM models, enzymatic, and process constraints with NIR spectroscopy-based measurements for feedback. It has been shown in the previous work that the maximization of HA production of a desired quality is linked to maintaining glucose and acetate concentrations in the extracellular environment at specific values. Hence, the objective function of the MPC problem will be formulated to maintain the desired setpoints of glucose and acetate concentrations. In the third objective, the MPC problem will be formulated to track setpoint changes in glucose and acetate concentration values for switching grades of HA in an online manner. The formulated MPC problems are bi-level optimization problems with the inner-loop optimization involving the RCM models and the outer-loop optimization involving maintaining the setpoints subject to the inner-loop optimization. The solution strategies to solve these bi-level optimization problems in an online manner will be developed by exploiting the structure of the problem. All the MPC strategies will be validated through a set of experiments.