Although retroviruses are a promising tool for gene therapy, there are two major problems limiting the establishment of viable industrial processes: retrovirus stability and low final yield in the supernatant. This fact emphasizes the need for an effective process optimization, not only at a genetic level but also at a bioprocess engineering level. In part 1 of this paper a mathematical model was developed to optimize the bioreaction yield by determining the best retrovirus harvest strategy in perfusion cultures. PA317 cells producing recombinant retroviruses were used to develop and test this model. Cell culture was performed in stirred tanks using porous supports. The parameters of the proposed model were experimentally determined for batch and perfusion cultures at 32 and 37 degrees C both with and without additives to enhance production; the model was then validated. This model allowed the determination of the optimal values of all operational variables included: batch and perfusion duration and perfusion rate. The highest productivity (2682 virus cm(-3) h(-1)) was obtained under the following conditions: batch at 37 degrees C for 53 h followed by perfusion at 32 degrees C for 23 h with a perfusion rate of 0.107 h(-1). This value was 3.5-fold higher than the best result obtained in batch cultures for the same conditions of titer and quality. A sensitivity analysis of the parameters showed that the parameters that affect most the final productivity depend on the bioreaction phase: cell growth in batch culture and production and product degradation in perfusion culture. In part 2 of this paper, this model is extended to the first step of downstream processing, and the addition of further steps to the process is discussed in order to achieve global process optimization.