Application of artificial neural network and fuzzy control for fed-batch cultivation of recombinant Saccharomyces cerevisiae

Abstract

A recombinant Saccharomyces cerevisiae expressing β-galactosidase under the control of the GAL10 promoter was constructed. The strain was used as a model to study fuzzy control with neural network, which served as state estimators for fed-batch cultivation of recombinant cells. To optimize the expression of β-galactosidase the effects of medium enrichment and induction on cell growth and expression were investigated. The activity of β-galactosidase was 2-fold higher in the presence compared with in the absence of yeast extract in the basal medium. Furthermore, the specific activity of β-galactosidase increased with increasing galactose concentration up to 30 g/ l. Two artificial neural networks (ANNs) were developed to estimate glucose and galactose concentrations using on-line measurements of ethanol and biomass concentrations, culture volume and the amount of carbon source fed to the fermentor. To improve productivity and product yield of β-galactosidase two multi-variable fuzzy controllers were used to control the glucose and galactose feed rates during the cell growth and production phases, respectively. Experimental data show that under fuzzy control with neural network estimators, the productivity was 2.7-fold higher than that in the case of exponential feeding, and 1.7-fold higher than that in the case of exponential feeding with feedback compensation using ethanol concentration.