Two Modeling Methods for Optimization of the Culture Conditions for Nisin Production by Lactococcus Lactis Subsp. Lactis

Two modeling methods were applied to optimize the culture conditions for nisin production by Lactococcus lactis subsp. lactis in shake flasks, respectively and the results obtained by these methods were compared. The effects of the candidate culture conditions on nisin titer (NTs) were investigated by Plackett-Burman design (PBD) experiments. A linear regression model was established using the data of PBD and the significant culture conditions were identified by F-test method. A box-behnken design was employed for further optimization. Using the data of the box-behnken design experiments, a quadratics regression model (QRM) and an artificial neural network (ANN) model was established for NTs prediction. The maximum NTs obtained by ANN model combined with Genetic algorithm (GA) was 25249.15 IU/ml which 1098.47 IU/ml higher than that obtained by QRM combined with derivative extreme method and 3826.15 IU/ml higher than that without optimization.