BACKGROUND: Response surface methodology is used to build a predictive model of the combined effects of independent variables (pH, salt concentration starch and casein). The model was validated in a laboratory-scale bioreactor for extracellular protease production from a newly isolated Serratia rubidaea. RESULTS: Optimum medium conditions obtained from the optimization experiments after 48 h incubation were starch, 8 g L-1; casein, 4 g L-1; salt concentration 6.25 g L-1; and initial pH, 8. A coefficient of determination (R 2) value of 0.9305 shows the fitness of the second-order model for the present studies. Results of model coefficients estimated by multiple linear regressions indicate that linear effects of casein concentration (P < 0.001308) and initial pH (P < 7.91 E-07) are more significant than similarly interactive effects of starch and casein (P < 0.019153), casein and salt concentration (P < 0.016294), casein and pH (P < 0.039904) and salt concentration and pH (p < 0.017845). The P-values of quadratic effects of casein, x2 x x2 (P < 0.000171); SC, x3 x x3 (P < 0.009134); initial pH, x4 x x4 (P < 0.000114) are more significant for maximal production. After optimization, protease production was enhanced experimentally by almost 65% in a shake flask and by almost 115% in a bioreactor. CONCLUSION: The alkaline proteases secreted by S. rubidaea were significant from an industrial perspective because of their stability against surfactants, oxidants and solvents. The statistical design is useful in economic protease production in a cost-effective medium for potential use on an industrial scale. (c) 2007 Society of Chemical Industry.