(656b) Grey-Box Stochastic Modelling of a an Enzymatic Reactions Network for Biotransformation

The increasing interest in producing complex fine chemicals and intermediates in the pharmaceutical industry using biochemical synthesis constitutes the general motivation behind the EUROBIOSYN project. Presently, only a few bio-transformations steps are involved in complex synthesis in industry, although enzymes are widely known as being specific, fast and working under mild conditions. To develop a purely enzymatic synthesis for complex molecules from simple (sugar) substrates, large reaction networks are necessary. One way to construct such a functional enzymatic reaction network is called System of Bio-transformations (SBT). A SBT is based on a part of one single organism's metabolic network containing the synthesis path including cofactor regeneration reactions. Mutants of E-coli microorganism are used to produce the suitably genetically modified metabolic network for an SBT. Thereby, the bio-transformations are performed as cell free extract in the production phase, combining the easy handling of a viable culture with the advantages of in vitro bio-transformations (Schümperli, 2006). The complexity of such large biochemical reaction networks involves a large number of reaction steps with many metabolites and enzymes, each playing different roles either as biocatalysts and/or as feed-forward and feedback regulators. For this presentation, the key product is Di-hydroxy-acetone phosphate (DHAP). DHAP is an important precursor for the production of phosphorylated, non-natural carbohydrates. Thereby, the DHAP-producing SBT contains all the enzymes for the glycolysis reactions, leading to a system of high complexity. The objective of the current work is to develop a methodology for optimizing a SBT for the desired production. In order to improve the yield of the system towards the desired product it is desirable to develop a model with capability of long-term prediction. To achieve this objective it is necessary both to develop and to apply an integrated methodology for identification, modeling and experimental design. Grey-box stochastic state space modeling representing a trade-off between theoretical and data knowledge is the methodology. The starting point is the grey-box stochastic modelling framework developed by (Kristensen et al, 2004), that will be further developed, and extended with a grey box stochastic experimental design feature. A simplified model of the network together with very limited experimental measurements represents the starting point in this project. The existing stochastic grey-box modelling framework was used to iteratively improve the very initial model by revealing some aspects of the enzyme kinetic. In essence an initial model structure in terms of a system of stochastic differential equations is formulated. The model parameters and unknown functional dependencies are iteratively estimated as the information in the experimental data is exploited. Using the model it is possible to separate the model mismatch and measurement errors. The model is reformulated and the iterations continued until the model is un-falsified using available data or all the information contained in the data with respect to the dynamics is exhausted. The presentation will focus on the grey-box stochastic modeling of SBT and will show the current results and progress.

Bibliography: 1. Kristensen N. R., Madsen H., Jørgensen S. B., 2004, ?A Method for Systematic Improvement of Stochastic Grey-Box Models?, Computers and Chemical Engineering, vol. 28, pp 1431-1449 2. Michael Schüperli, Matthias Heinemann, Anne Kümel, and Sven Panke, in preparation, 2006