Abstract:  For improving the poly-γ-glutamate (PGA) yield, the orthogonal design was used for the trial design of the formula of medium components: glutamate, glucose, citrate and glycerol, radius basis function neural network (RBFNN) was applied for the predict modeling of the relationships between the PGA yield and the concentration of medium components. Then the genetic algorithm (GA) was used for the global optimization of the model. The optimum combination of the medium was obtained: glutamate 21.2g/L, glucose 75.4g/L, citrate 7.2g/L, glycerol 10.8g/L. The yield of PGA was improved to 12.8g/L, which was increasedby 39.1 % compared to the original medium.

Key words:  , poly-&gamma, -glutamate； RBF neural network； genetic algorithm； fermentation medium； optimization；