Biological production processes are far less well understood than chemical ones, and therefore the need for model-based control and optimization is even greater in this domain. However, the complexity of the biological processes makes modeling very challenging, and models of adequate complexity must be formulated that on the one hand represent the biological knowledge as much as possible but on the other hand do not have too many uncertain parameters.
Modeling and control of yeast fermentations
Saccharomyces cerevisiae, also known as budding yeast, is widely used in the food and beverage industry, in particular for baking, winemaking, and brewing. The early work of the dyn group focused on modeling of yeast cell cultures with respect to the cell cycle and on developing strategies to synchronize the budding of the cells. Currently, we investigate the transition of yeast cells to ethanol production and how to avoid this transition while achieving maximal growth rates. This phenomenon is called the Crabtree effect after its discoverer. The work is part of the BMBF project „YeastScent” in which ion mobility spectrometry (IMS) is used for the measurement of volatile metabolites in the off-gas of yeast fermentations and for the control of the feeding strategy. Our group develops a suitable mathematical model to predict the switching to ethanol production. The model is based on the biochemistry of the cell and extended by kinetics expressions.
Publications
Wegerhoff, S., Neymann, T.C., Engell, S.: Synchronization of a budding yeast cell culture by manipulating inner cell cycle concentrations. In: Proc. of the 2012 IEEE Decision and Control Conference (CDC), Maui, USA, 2012.
Wegerhoff, S., Engell, S.: Simulation of the aerobic growth of Saccharomyces cerevisiae during fed batch fermentation by dynamic flux balance analysis. In: Proc. of the 2015 Foundation of System Biology in Engineering (FOSBE), Boston, USA, 2015.
Dynamic models of biosystems based upon elementary modes
Mathematical models can support the development and the operation of bioprocesses significantly, by their use in model-based design of experiments to find optimal operating conditions, in optimizing batch trajectories, and in monitoring and control strategies. Recent advances in process-analytical technology provide access to a large amount of offline and online data of fermentation processes, but utilization of this data requires mathematical process models which characterize the important dependencies with a limited model complexity but sufficient accuracy.
In a project with Bayer Technology Services, we are developing a systematic modeling procedure so that dynamic process models can be generated within a short period of time. The procedure uses both knowledge-based and data-based modeling techniques and can be applied to batch- and fed batch processes. Elementary Modes are used as representative macro reactions for the cell metabolism. A newly developed data reconciliation procedure is used to test the importance of Elementary Modes. This method is further used in a multi-objective optimization for the selection of an ideal set of macro reactions. The influences of the process conditions are analyzed statistically and integrated in the kinetic expressions for the reaction rates. This approach utilizes the available biochemical knowledge and thus goes beyond the usual approach based only upon formal kinetics that depend on external concentrations and still leads to fast model development. The generated process models represent the process well within the design space and can be applied for online applications as e.g. state estimation and nonlinear model-predictive control (NMPC) at the real process.
Publication
Hebing, L., Neymann, T.C., Engell, S.: An Efficient Modelling Procedure for the Generation of Bioprocess Models with adaptable Complexity. In: Proc. of the 2015 Foundation of System Biology in Engineering (FOSBE), Boston, USA, 2015.