3.4. Response surface methodology
The significant independent variables (glycerol, MgSO47H2O, initial pH) were used to determine the optimum levels of these parameters based on the above results. The Box-Behnken design with three factors and three levels, including five replicates at the centre point, was used for fitting a second-order response surface (Table 4). The design matrix and the corresponding experimental data were shown in Table 4. The results were analyzed by standard analysis of variance (ANOVA) and following quadratic regression equations were obtained in terms of protease production. Using the designed experimental data, the polynomial model for protease yield Y was regressed by only considering the significant terms and was shown as below