This paper proposes a general approach for building a mechanistic yeast model able to predict the shift of metabolic pathways. The mechanistic model accounts for the coexistence of several metabolic pathways (aerobic fermentation, glucose respiration, anaerobic fermentation and ethanol respiration) whose activation depends on growth conditions. This general approach is applied to a commercial strain of Saccharomyces cerevisiae. Stoichiometry and yeast kinetics were mostly determined from aerobic and completely anaerobic experiments. Known parameters were taken from the literature, and the remaining parameters were estimated by inverse analysis using the particle swarm optimization method. The optimized set of parameters allows the concentrations to be accurately determined over time, reporting global mean relative errors for all variables of less than 7 and 11% under completely anaerobic and aerobic conditions, respectively. Different affinities of yeast for glucose and ethanol tolerance under aerobic and anaerobic conditions were obtained. Finally, the model was successfully validated by simulating a different experiment, a batch fermentation process without gas injection, with an overall mean relative error of 7%. This model represents a useful tool for the control and optimization of yeast fermentation systems. More generally, the modeling framework proposed here is intended to be used as a building block of a digital twin of any bioproduction process.