Using available biochemical information, metabolic networks have been constructed to describe the biochemistry of growth of Saccharomyces cerevisiae and Candida utilis on a wide variety of carbon substrates. All networks contained only two fitted parameters, the P/O ratio and a maintenance coefficient. It is shown that with a growth-associated maintenance coefficient, K, of 1.37 mol ATP/ C-mol protein for both yeasts and P/O ratios of 1.20 and 1.53 for S. cerevisiae and C. utilis, respectively, measured biomass yields could be described accurately. A metabolic flux analysis of aerobic growth of S. cerevisiae on glucose/ethanol mixtures predicted five different metabolic flux regimes upon transition from 100% glucose to 100% ethanol. The metabolic network constructed for growth of S. cerevisiae on glucose was applied to perform a theoretical exercise on the overproduction of amino acids. It is shown that theoretical operational product yield values can be substantially lower than calculated maximum product yields. A practical case of lysine production was analyzed with respect to theoretical bottlenecks limiting product formation. Predictions of network-derived irreversibility limits for Y(sp) (mu) functions were compared with literature data. The comparisons show that in real systems such irreversibility constraints may be of relevance. It is concluded that analysis of metabolic network stoichiometry is a useful tool to detect metabolic limits and to guide process intensification studies. (c) 1995 John Wiley & Sons, Inc.