Optimization of fermentation medium for nisin production from Lactococcus lactis subsp. lactis using response surface methodology (RSM) combined with artificial neural network-genetic algorithm (ANN-GA)
Nisin is a bacteriocin approved in more than 50 countries as a safe natural food preservative. Response surface methodology (RSM) combined with artificial neural network-genetic algorithm (ANN-GA) was employed to optimize the fermentation medium for nisin production. Plackett-Burman design (PBD) was used for identifying the significant components in the fermentation medium. After that, the path of steepest ascent method (PSA) was employed to approach their optimal concentrations. Sequentially, Box-Behnken design experiments were implemented for further optimization. RSM combined with ANNGA were used for analysis of data. Specially, a RSM model was used for determining the individual effect and mutual interaction effect of tested variables on nisin titer (NT), an ANN model was used for NT prediction, and GA was employed to search for the optimum solutions based on the ANN model. As the optimal medium obtained by ANN-GA was located at the verge of the test region, a further Box- Behnken design based on the RSM statistical analysis results was implemented. ANN-GA was implemented using the further Box-Behnken design data to locate the optimum solution which was as follow (g/l): Glucose (GLU) 15.92, peptone (PEP) 30.57, yeast extraction powder (YEP) 39.07, NaCl 5.25, KH2PO4 10.00, and MgSO4·7H2O 0.20, with expected NT of 22216 IU/ml. The validation experiments with the optimum solution were implemented in triplicate and the average NT was 21423 IU/ml, which was 2.13 times higher than that without ANN-GA methods and 8.34 times higher than that without optimization.
Key words: Response surface methodology, artificial neural network, genetic algorithm, nisin titer.