CONICET | Buscador de Institutos y Recursos Humanos

Models of cultures of genetically modified microorganisms are widely used for deepening the understanding of intracellular processes and their interactions with bioreactor operating conditions that influence productivity and performance. More specifically, optimization of biomass production by integrating imperfect models with data is key to increase the amount obtained of a desired end-product product such as a protein. In this work, a model-based optimization methodology that uses experimental feedback is applied to a fed-batch bioreactor. Typically, model-based optimization approaches may have acceptable convergence rates to a local optimum, but they are negatively affected by modeling errors when extrapolating to unknown operating conditions. Experimental feedback is used here to solve this problem. After the model has been (re)parameterized, an optimized experiment is designed to maximize the performance of the bioprocess. Data gathered in this experiment is used to correct the model, and the cycle continues until no further improvement is found. The method is tested in the production of baker´s yeast biomass, which is a challenging problem due to the Crabtree effect. After a model is proposed and analyzed, several successive experiments are designed and performed in a bench-scale bioreactor. Results obtained demonstrate the capability of the proposed approach to find an improved feeding profile that leads to better performance with minimum experimental effort.