This paper describes improved optimization method that combines the Plackett-Burman design, a full factorial design and the response surface method, which were used to optimize the medium for the production of rifamycin B by Amycolatopsis mediterranei S699. The five nutrients used to amend the medium were preliminary selected from 15 medium components based on the results of the Plackett-Burman design experiments and subsequent screening process using the full factorial design experiments was performed. Glucose, (NH(4))(2)SO(4), and corn steep liquor were found to be the best nutrients for the optimization of rifamycin B productions, and central composite design experiments indicated that the optimal concentrations of these nutrients were 63 g/L, 7.9 g/L, and 10 g/L, respectively. A bioreactor culture was used to investigate the production characteristics obtained when Amycolatopsis mediterranei S699 was grown in the determined optimum medium. Although cell growth stopped at 88 h, rifamycin B production continued, reaching its maximum value at the end of the experiment (172 h) after the glucose in the medium was completely consumed. The maximum cell concentration and rifamycin B titer were 8.6 g/L and 1260 mg/L, respectively. The combined optimization method described here is new and effective method for screening carbon and nitrogen sources as well as determining their optimum levels. (C) 2008 Elsevier Ltd. All rights reserved.