Comparison of artificial neural network (ANN) and response surface methodology (RSM) in optimization of the immobilization conditions for lipase from Candida rugosa on Amberjet(®) 4200-Cl.

Benamia Fatiha, Bouchagra Sameh, Saihi Youcef, Djeghaba Zeineddine, Rebbani Nacer

Index: Prep Biochem Biotechnol. 43(1) , 33-47, (2013)

Full Text: HTML

Abstract

Candida rugosa lipase (CRL) is an important industrial enzyme that is successfully utilized in a variety of hydrolysis and esterification reactions. This work describes the optimization of immobilization conditions (enzyme/support ratio, immobilization temperature, and buffer concentration) of CRL on the anionic resin Amberjet® 4200-Cl, using enantioselectivity (E) as the reference parameter. The model reaction used for this purpose is the acylation of (R,S)-1-phenylethanol. Optimal conditions for immobilization have been investigated through a response surface methodology (RSM) and artificial neural network (ANN). The coefficient of determination (R(2)) and the root mean square error (RMSE) values between the calculated and estimated responses were respectively equal to 0.99 and 0.06 for the ANN training set, 0.97 and 0.2 for the ANN testing set, and 0.94 and 0.4 for the RSM training set. Both models provided good quality predictions, yet the ANN showed a clear superiority over RSM for both data fitting and estimation capabilities.