Effect of experimental error on the efficiency of different optimization methods for bioprocess media optimization

Abstract.

Four optimization methods (Simplex, Rosenbrock, iterative factorial experimental design (IFED) and genetic algorithms) for the optimization of the biotechnological media composition under conditions where the measured quantities are subjected to the experimental error were compared. The computer simulations were performed on some of the selected two- to six- parameter biotechnological models. The optimization process was modified in such a way that the experimental error was considered. The results show that the optimization efficiency increases when this new termination criteria is implemented. In addition, the method efficiency becomes independent of the experimental error. In general, Simplex and Rosenbrock methods need fewer experiments and their distribution of necessary experiments is narrower than for IFED and genetic algorithms. The increase of model parameters that need to be optimized results in a decrease in the method efficiency and in an increase of the average number of required experiments. The results were further verified on the cultivation of Saccharomyces cerevisiae and were found to be in good agreement with the results obtained from the computer simulations.