Optimization of medium constituents for Cephalosporin C production using response surface methodology and artificial neural networks

Artificial neural networks (ANN) and response surface methodology(RSM) were used to build a model to describe the effects of fourindependent variables (moisture content, concentrations of glucose,ammonium nitrate and methionine) on the yield of cephalosporin C(CPC) from Acremonium chrysogenum under solid statefermentation. The respective uses of RSM and ANN were found tobe effective in locating the optimum conditions within the rangefixed from the preliminary runs. When compared with thepredictions given by RSM, ANN was found to be superior indescribing the fermentation process for the production of CPC.When a global optimization routine was employed to optimize theequation resulted from the neural networks, the optimum predictedantibiotic yield was found to be 29.4 mg/g which is 14.8 % higherthan the optimum value obtained from preliminary runs, and 9.2 %higher than value obtained from Box-Behnken design of RSM.