Bioprocesses are nonlinear and time varying in nature, and reliable modeling techniques ate needed for their control and optimization. In this work, the performance of several input-output models is investigated in a recombinant fed-batch fermentation using a series of statistical methods. An identification experiment was designed and implemented to provide the best system response to generate the process. Statistical analyses sere performed to characterize the information. Several criteria were defined to compare the different models, Use of this rigorous modeling methadology allowed objective comparison of model performance, A detailed methodology is presented for input-output modeling based on statistical analysis, Nonlinear models with constant parameters and linear models,vith time varying parameters were the most appropriate models for the case studied, but the choice of model will depend on the ultimate control or optimization objectives.