Avermectin B1b production optimization from Streptomyces avermitilis 41445 UV 45(m)3 using response surface methodology and artificial neural network

Abstract

Present study was conducted to optimize avermectin B1b production from S.avermitilis 41445 UV45(m)3 using artificial neural network and response surface methodology. Three variables NaCl, KCl, and pH were used for optimization. Coefficient of determination and adjusted coefficient of determination have very poor values for RSM. Values predicted by RSM for experiments were also much different from the observed avermectin production. Comparatively predicted avermectin levels by ANN were very close to observed values with much higher R2 and adjusted R2. Optimum levels of NaCl, KCl, and pH predicted by ANN were 1.0 g/L, 0.5 g/L, and 7.46 respectively. Sensitivity analysis predicted highest effect being shown was by pH followed by NaCl and KCl. About 37.89 folds increase in avermectin B1b production was observed at optimum levels of three variables envisage by ANN. Optimum levels, ranking order of variables, and the predicted avermectin on the optimum levels by the RSM was much different from ANN values. Results revealed that ANN is a better optimization tool for given strain than RSM. © 2014 The Korean Society for Applied Biological Chemistry.