Most cybernetic modeling efforts to date have taken a minimalist viewpoint of metabolism, relying heavily upon process lumping and pathway abstraction to simplify the underlying reaction network. However, such models are unsuitable for metabolic engineering applications because they do not incorporate the level of biological detail that is needed to predict the effects of genetic perturbation. In this paper, we inquire into the formulation and analysis of structured cybernetic models as tools for metabolic engineering. We present fresh theoretical developments leading to a more systematic treatment of global cybernetic variables based on conservation of metabolic resources. Finally, we illustrate these concepts with a cybernetic model-based analysis of anaerobic Escherichia coli metabolism. The model is able to accurately describe the growth of wild-type and acetate knockout strains when fitted to experimental data. Although it adequately predicts the end-of-batch concentrations for a strain with intermediate pta-ackA activity, it overestimates the measured growth and glucose uptake rates. (C) 2004 Elsevier Ltd. All rights reserved.