Evaluation of alternative model structures of metabolic systems: two case studies on model identification and validation

In the paper some quantitative criteria for evaluating alternative linear model structures are presented and applied to two metabolic systems. Linear time-invariant structured models and input/output descriptions are proposed for the two metabolic processes considered, bilirubin and riphamicin kinetics, and maximum likelihood parameter estimates are obtained from input/output test data. The considered criteria include model plausibility and quantitative measures, e.g. the F-test, Akaike information criterion and the inverse of the information matrix, which are based on the results of model identification from input/output data. The combined use of the Akaike criterion and of the inverse of the information matrix proves to be, in conjunction with model plausibility, a very reliable tool in the metabolic model structure determination.

[1]  Y. Z. Ider,et al.  Quantitative estimation of insulin sensitivity. , 1979, The American journal of physiology.

[2]  Kiyoshi Yamaoka,et al.  Application of Akaike's information criterion (AIC) in the evaluation of linear pharmacokinetic equations , 1978, Journal of Pharmacokinetics and Biopharmaceutics.

[3]  C. Cobelli,et al.  Parameter and structural identifiability concepts and ambiguities: a critical review and analysis. , 1980, The American journal of physiology.

[4]  Mario Milanese,et al.  Methods for the Selection of Approximating Classes of Models , 1979 .

[5]  R. Brown,et al.  Adequacy of measurements in compartmental modelling of metabolic systems , 1979, Medical and Biological Engineering and Computing.

[6]  N I Berlin,et al.  Studies of bilirubin kinetics in normal adults. , 1969, The Journal of clinical investigation.

[7]  L Finkelstein,et al.  Modeling and identification of metabolic systems. , 1981, The American journal of physiology.

[8]  C. Cobelli,et al.  A modeling study of the effect of fasting on bilirubin kinetics in Gilbert's syndrome. , 1981, The American journal of physiology.

[9]  H. Akaike Fitting autoregressive models for prediction , 1969 .

[10]  H. Akaike,et al.  Information Theory and an Extension of the Maximum Likelihood Principle , 1973 .

[11]  R. Fletcher A modified Marquardt subroutine for non-linear least squares , 1971 .

[12]  A. Kelman,et al.  Modeling of drug response in individual subjects , 1980, Journal of Pharmacokinetics and Biopharmaceutics.

[13]  K. Tranberg,et al.  Modeling of plasma disappearance of unlabeled insulin in man. , 1978, The American journal of physiology.

[14]  Karl Johan Åström,et al.  On the Achievable Accuracy in Identification Problems , 1967 .

[15]  R. Wolfe,et al.  The reliability of rates of glucose appearance in vivo calculated from single tracer injections. , 1979, Canadian journal of physiology and pharmacology.

[16]  T. Söderström On model structure testing in system identification , 1977 .