Genetically modified Saccharomyces cerevisiae strain (YPB-G) which secretes a bifunctional fusion protein that contains both Bacillus subtilis alpha-amylase and Aspergillus awamori glucoamylase, activities was used for the direct conversion of starch into ethanol. Starch was either supplied initially to different nutrient media or added instantaneously to the reactor at various discrete time instants (pulse feeding). Stoichiometric modeling was used to investigate the effects of initial substrate concentration and growth rate of the recombinant yeast culture on ethanol production. Reaction stoichiometries describing both the anabolism and catabolism of the microorganism were used as an input to flux balance analysis (FBA), the preferred metabolic modeling approach since the constructed stoichiometric network was underdetermined. Experiments for batch and fed-batch systems at different substrate concentrations were analyzed theoretically in terms of flux distributions using ethanol production rate as the maximization criteria. Calculated ethanol rates were in agreement with experimental measurements, suggesting that this recombinant microorganism is sufficiently evolved to optimize its ethanol production. The function of the main pathways of yeast metabolism (PPP, EMP, TCA) are discussed together with the node analyses of glucose-6-P and pyruvate branch points. Theoretical node analysis revealed that if the split ratio in G6P branch point is changed by genetic manipulations, the ethanol yield would be affected considerably. (C) 2003 Elsevier Ltd. All rights reserved.