On the development of an intelligent control system for recombinant cell culture

Abstract

To produce a recombinant protein, it is critically important to optimize and control bioprocesses based on knowledge of a cell's genetic, metabolic, and kinetic behavior. It is, however, not straightforward due to the fact that the biosystem is highly nonlinear, time variant, and complex. In this paper, we discuss three kinds of intelligent control systems for control of fed-batch cultivation of recombinant E. coli and yeast, namely, fuzzy pH-stat, fuzzy neural network, and fuzzy control coupled with a neural network estimator. In a fuzzy pH-stat control system, the relationship between pH change in the medium and glucose consumption rate is modeled by a fuzzy set and thereafter used to control the feed rate of glucose to obtain cell density as high as 72 g/L. In a fuzzy neural network (FNN) control system, a FNN was constructed to learn fuzzy control inference and then was applied to fed-batch cultivation of recombinant E. coli to attain a high expression of recombinant protein. In addition, we developed a fuzzy control system coupled with neural network estimators that can on-line estimate residual glucose and galactose concentrations, which were utilized to control the feed rate of glucose (during the cell growth phase) and the feed rate of galactose (during the expression phase). Results of these control strategies are presented and their usefulness in the fed-batch cultivation of recombinant strains is demonstrated. The idea behind these studies is to utilize predetermined experimental data to develop repetitive learning control using intelligent techniques.