A metabolic network model is one of the cornerstones of the emerging Metabolic Engineering methodology. In this article, special attention is therefore, given to the phase of model building. A five-stage structured approach to metabolic network modeling is introduced. The basic steps are : (1) to collect a priori knowledge on the reaction network and to build candidate network models, (2) to perform an a priori check of the model, (3) to estimate the unknown parameters in the model, (4) to check the identified model for acceptability from a biological and thermodynamic point of view, and (5) to validate the model with new data. The approach is illustrated with a growth system involving baker＇s yeast growing on mixtures of substrates. Special attention is given to the central uncertainties in metabolic network modeling, i.e., estimation of energetic parameters in the network and the choice of the source of anabolic reducing equivalents NADPH.