Modeling of metabolic networks and flux analysis
Mathematical process models are an important cornerstone for analyzing, optimizing and controlling a bioprocess in a reliable way. For this purpose, modeling of metabolic networks a is very important task. To characterize such processes precisely, detailed models of relevant sections of the metabolic network, the kinetics of all reactions involved and knowledge about metabolic regulations are essential. For developing metabolic models, methods of metabolic flux analysis in combination with algorithms for building elementary flux models are used. These models are extended to Cybernetic Models by an unified description of rate-determining reactions, including kinetics of possibly rate limiting reactions, and of dynamics of metabolic regulations. Cybernetic modelling based on the assumption that the coordination of metabolic reaction follows an optimal criterion is often most efficient. Until now, models for cultivating baker's yeast, Escherichia coli and Penicillium as well as models for cultivating animal cells have been implemented and used for process control and optimization.
Expert systems and model based procedures
One focus in this field of research lies on computer based identification of disruptions in sensor and actuator operations (such as failure or drift) and on identification of variations in biological systems. This is for example caused by "non reproducible" inoculation conditions, changes in substrate quality or infections. For this purpose an expert system is established which collects experience of the operating personnel. This expert system is complemented with a-priori knowledge of the process (theoretical models) and methods of time series analysis to carry out the measurements as accurately as possible. In case of unexpected changes, the expert system can use heuristic control strategies to bring the process back into a range where it can be handled by model based control strategies again. The Extended Kalman Filter is used for estimation of state variables which can also be employed for feed-forward and feedback control of substrate concentration during fed-batch cultivations. This type of control attains the highest possible specific growth rate under strict aerobic metabolism at low substrate concentrations (e.g. glucose concentration of 0.1 g/l and lower) so that no toxic or unwanted metabolites (acetate with bacteria, ethanol with yeast, lactate with animal cells) are produced.
Analysis and optimization of processes
To extract unknown process information from raw data, data-mining strategies are used. Reaction kinetics and transport processes are described with the help of mathematical models to carry out simulations for gathering insights about the process's dynamics. For example, optimal feeding rates for fed-batch processes guaranteeing high space-time-yield are computed using process models. Furthermore, optimized design of experiment strategies are used to determine optimal sampling times and to minimize estimation error covariances when estimating process parameters.