A metabolic flux analysis (MFA) model was developed to optimize the xylose conversion into ethanol using Candida shehatae strain. This metabolic model was compartmented and constructed with xylose as carbon substrate integrating the enzymatic duality of the first step of xylose degradation via an algebraic coefficient. The model included the pentose phosphate pathway, glycolysis, synthesis of major metabolites like ethanol, acetic acid and glycerol, the tricarboxylic acid cycle as well as the respiratory chain, the cofactor balance, and the maintenance. The biomass composition and thus production were integrated considering the major biochemical synthesis reactions from monomers to each constitutive macromolecule (i. e., proteins, lipids, polysaccharides, nucleic acids). The construction of the model resulted into a 122-linear equation system to be resolved. A first experiment allowed was to verify the accuracy of the model by comparing calculated and experimental data. The metabolic model was utilized to determine the theoretical yield taking into account oxido-reductive balance and to optimize ethanol production. The maximal theoretical yield was calculated at 0.62 Cmol(ethanol)/Cmol(xylose) for an oxygen requirement of 0.33 moloxygen/molxylose linked to the cofactors of the xylose reductase. Cultivations in chemostat mode allowed the fine tuning of both xylose and oxygen uptakes and showed that lower was the oxygen/xylose ratio, higher was the ethanol production yield. The best experimental ethanol production yield (0.51 Cmol(ethanol)/Cmol(xylose)) was obtained for an oxygen supply of 0.47 moloxygen/molxylose.