A hybrid input-output approach to model metabolic systems: An application to intracellular thiamine kinetics

Models of the dynamics of complex metabolic systems offer potential benefits to the deep comprehension of the system under study as well as for the performance of certain tasks. Unfortunately, dynamic modeling of a great deal of metabolic systems may be problematic due to the incompleteness of the available knowledge about the underlying mechanisms and to the lack of an adequate observational data set. In theory, a valid alternative to classical structural modeling through ordinary differential equations could be represented by input-output approaches, But, in practice, such methods, which learn the nonlinear dynamics of the system from input-output data, fail when the experimental data set is poor either in size or in quality. Such a situation is not rare in the case of metabolic systems. This paper deals with a hybrid approach which aims at overcoming the problems addressed above. More specifically, it allows us to solve the identification problems of the intracellular thiamine kinetics in the intestine tissue. The method, which is half way between the structural and input-output approach, uses the outcomes of file simulation of a qualitative structural model to build a good initialization of a fuzzy system identifier. Such an initialization allows us to efficiently cope with both the incompleteness of knowledge and the inadequacy of the available data set, and to derive an input-output model of the intracellular thiamine kinetics in the intestine tissue. The comparison of the predictions of the intracellular thiamine kinetics obtained by the application of such a model with those obtained by traditional approaches, namely compartmental models, neural networks, and fuzzy systems, highlighted a better performance of our model. As the structural assumptions are relaxed, we obtained a model slightly less informative than a purely structural one but robust enough to be used as a simulator. The paper also discusses the interpretative potential offered by such a model, as tested on diabetic subjects. (C) 2001 Elsevier science (USA).