Parameter estimation in kinetic reaction models using nonlinear observers facilitated by model extensions

Tools

Fey, D. and Findeisen, R. and Bullinger, E.; Chung, Myung Yin and Misar, Pradeep, eds. (2008) Parameter estimation in kinetic reaction models using nonlinear observers facilitated by model extensions. In: Proceedings of the 17th IFAC World Congres. IFAC Proceedings, 41 . International Federation of Automatic Control, Seoul, Korea, pp. 313-318. ISBN 9783902661005 (https://doi.org/10.3182/20080706-5-KR-1001.00053)

[thumbnail of strathprints007337]
Preview
 Text. Filename: strathprints007337.pdf
Accepted Author Manuscript
Download (155kB)| Preview

Abstract

An essential part of mathematical modelling is the accurate and reliable estimation of model parameters. In biology, the required parameters are particularly difficult to measure due to either shortcomings of the measurement technology or a lack of direct measurements. In both cases, parameters must be estimated from indirect measurements, usually in the form of time-series data. Here, we present a novel approach for parameter estimation that is particularly tailored to biological models consisting of nonlinear ordinary differential equations. By assuming specific types of nonlinearities common in biology, resulting from generalised mass action, Hill kinetics and products thereof, we can take a three step approach: (1) transform the identification into an observer problem using a suitable model extension that decouples the estimation of non-measured states from the parameters; (2) reconstruct all extended states using suitable nonlinear observers; (3) estimate the parameters using the reconstructed states. The actual estimation of the parameters is based on the intrinsic dependencies of the extended states arising from the definitions of the extended variables. An important advantage of the proposed method is that it allows to identify suitable measurements and/or model structures for which the parameters can be estimated. Furthermore, the proposed identification approach is generally applicable to models of metabolic networks, signal transduction and gene regulation.

CORE (COnnecting REpositories)