Abstract: The response surface methodology(RSM)was employed to optimize the medium composition for Epothilone B produced by Sorangium cellulosum SoF5-76. On the basis of one-variable-at-a-time design, using Plackett-Burman design, potato starch, skimmed milk powder and calcium chloride were screened out of 9 factors as main affecting variables of Epothilone B production, and then steepest ascent method was used to approach their maximum response regions, followed by Box-Behnken design, multiple regression analysis and response surface analysis. The optimum medium formula determined was composed of potato starch 3.9 g/L, skim milk powder 2.2 g /L, chlorine calcium 1.3 g/L, glucose 1 g/L, soybean powder 1.5 g/L, magnesium sulfate 2.5 g/L, EDTA-Fe3 + 3 mL/L, trace elements 0.5 mL/L, VB12 1 mL/L. Under this optimal conditions, Epothilone B production reached up to 29.95 mg/L and increased 1.1 times in average as compared with preliminary culture and consistent with the maxmum predicted value.