MODELLING AND OPTIMIZATION OF FERMENTATIVE MEDIUM FOR THE β-GALACTOSIDASE PRODUCTION FROM Kluyveromyces marxianus USING ARTIFICIAL NEURAL NETWORK (ANN)

β-galactosidase (β-gal) has considerable interest in various biotechnological applications and its production is influenced by many factors that need to be optimized. Recently, modelling machine learning techniques such as artificial neural networks (ANNs) have gained attention. The objectives of this study were model and optimize the fermentative medium for the β-gal production from Kluyveromyces marxianus (ATCC 46537) using artificial neural network (ANN). Submerged fermentation (erlenmeyer) containg lactose (whey permeate), yeast extract and (NH4)2SO4 was used to produce K. marxianus β- gal. The feed-forward, fully-connected ANN trained with the Back-propagation algorithm was the tool used for get the model. The optimization was done using the differential evolution algorithm. The specific activity of the K. marxianus β-gal ranged from 15.16 to 35.92 U mgprotein-1 The highest observed value was obtained at the central point of the CCD. ANN performance was evaluated by MSE and R2, which showed good agreement with experimental results. The optimal condition founded was almost 20% higher than the best value found in the experiments of the CCD, indicating the importance of finding optimum production conditions, in order to maximize productivity and reduce the costs.