Artificial intelligence based optimization of fermentation medium for β-glucosidase production from newly isolated strain Tolypocladium cylindrosporum

Abstract

A Tolypocladium cylindrosporum strain was isolated for efficiently produce extracellular thermoacidophilic β-glucosidase (BGL). This objective of the present paper is to integrate two different artificial intelligence techniques namely artificial neural network(ANN) and genetic algorithm(GA) for optimizing medium composition for the production of BGL on submerged fermentations(SmF). Specifically, the ANN and GA were used for modeling non-linear process and optimizing the process. The experimental data reported in a previous study for statistical optimization were used to build the ANN model. The concentrations of the four medium components served as inputs to the ANN model and the β-glucosidase activity as the output of the model. The average error (%) and correlation coefficient for the ANN model were 1.36 and 0.998, respectively. The input parameters of ANN model were subsequently optimized using the GA. The ANN-GA model predicted a maximum β-glucosidase activity of 2.679U/ml at the optimun medium composition. The ANN-GA model predicted gave a 22% increase of β-glucosidase activity over the statistical optimization, which was in good agreement with the actual experiment under the optimum conditions. © 2010 Springer-Verlag.