An on-line methanol sensor system was developed using a methanol probe, methanol sensor unit and peristaltic pump. The system was commanded using data acquisition (DAQ) and LabVIEW software. Calibration of the methanol sensor system was done in a medium environment with yeast cells during cells adaptation to methanol metabolism after glycerol feeding was stopped. The correlation equations between voltage output signal from the methanol sensor unit and residual methanol in culture broth were created with third order polynomial regression. This developed system was implemented for online methanol control in recombinant human serum albumin (rHSA) protein production by P. pastoris KM71 at methanol levels of 4 and 10 g/l with controlled fluctuations at 13.0 and 11.3% of oscillation, respectively. The accumulated amounts of recombinant protein from two levels of methanol concentration controls (4 and 10 g/l) were similar but the proteins were produced at a different rate related with methanol concentration in the broth. Therefore, the control at 10 g/l methanol had a higher production rate (0.53 mg-protein/g dry-cellh) than 4 g/l methanol control (0.38 mg-protein/g dry-cellh) as it reached the maximum protein concentration in a shorter time, even though its cell yield was less than that of 4 g/l methanol control. At the end of the experiments, the high cell density environment caused both cell and protein reduction by cell autolysis and protease degradation. However, the protein decrease could be prevented by taking protein induction at a low temperature and a pH where protease does not function.

In order to describe and predict the growth and expression of recombinant proteins by using a genetically modified Pichia pastoris, we developed a number of unstructured models based on growth kinetic equation, fed-batch mass balance and the assumptions of constant cell and protein yields. The growth of P. pastoris on both glycerol and methanol could be represented by Monod kinetic equation. A simple simulation methodology and developed models were shown to satisfactorily describe both growth and production of recombinant human serum albumin (rHSA) using a genetically modified P. pastoris Mut S strain. The obtained parameters from curve fitting were reasonable and could be acceptable. Moreover, the same parameter sets obtained by the experiments indicated the rigidity and consistency of the developed models and fermentation approach of this study. With correlation coefficients (r 2 ) exceeding 0.99, the models were able to simulate and predict the cell growth behavior and recombinant protein production by P. pastoris without requiring complex models.