Vous êtes ici : GIPSA-lab > Animation > Séminaires

Séminaire du département Automatique du 04/04/2013 à 14h00


On identifiability and identification of metabolic network models.

Intervenant : E. Cinquemani, Inria, Montbonnot (France)

Lieu : Salle Master Département Auto., ENSE3, Bat 3, 2è étage


Résumé :

The metabolism of a cell is the family of chemical reactions that govern the transformation of substrates
 (e.g. sugars)  into products (ATPs, aminoacids, etc.) necessary for the functioning of the cell. 
Mathematical modeling is thus essential  for understanding cellular physiology and for engineering modified
 cell functions.
High-throughput measurement techniques provide a wealth of information for the identification of metabolic
 network models,  but the problem remains challenging. Parameter identifiability may not hold due to 
structural modelling issues, or to limitations  in the observation of the system. In addition, measurement
errors and missing data entries result in practical identifiability issues and make data difficult to exploit at full.
Finally, metabolic network models are typically large and nonlinear, which complicates analysis and
numerical optimization. In this talk, I will discuss identifiability analsys and methods for the identification of 
metabolic network models. I will introduce formal statements of structural and practical identifiability and 
establish links among these properties. Based on this, I will discuss model reduction methods, and show
results from the application of these methods to data from the Escherichia coli central carbon metabolism.
Then, I will address estimation of the model parameters, with emphasis on methods for coping with incomplete
datasets and for improving inference of the system dynamics.

GIPSA-lab, 11 rue des Mathématiques, Grenoble Campus BP46, F-38402 SAINT MARTIN D'HERES CEDEX - 33 (0)4 76 82 71 31