Effect of carbon, nitrogen and thiamine HCl on endoglucanase production by Trichoderma viride: optimization and prediction model development

Trichoderma species are widely for the commercial production of cellulolytic enzymes. In the present investigation, cellulose (as carbon), soy peptone (nitrogen) and thiamine HCl (as vitamin) were optimized using response surface methology (RSM) and artificial neural network (ANN) to produce endoglucanase (EG) from Trichoderma viride. In RSM model, three factors were involved in Box-Behnken design. The EG production predicted by the ANN consisted of one output and three input layer neurons. Based on the coefficient of determination (R۲) and mean absolute error (MAE), RSM model provided a good quality prediction for the EG production in terms of all three variables. The optimum number of hidden neurons for ANN was found to be ۱۰ when the lowest values of mean squared error (MSE) for training, testing, and validation were determined. RSM and ANN techniques predicted the highest EG (۱.۴۵ U/ml) production with ۱.۸۸% cellulose, ۰.۵۵% soy peptone, and ۰.۱% thiamine HCl. EG production in the optimized medium was ۱.۷ fold higher than in unoptimizes medium. Appraisal of the models through the R۲, MAE and MSE showed that the ANN was superior to the RSM model in predicting the EG production.

Keywords:

artificial neural network , enzyme , response surface methodology , Trichoderma