Modeling and optimization III: Reaction rate estimation using artificial neural network (ANN) without a kinetic model

Abstract In this study, the usability of artificial neural networks (ANN) for the estimation of enzymatic reaction rate was investigated. The study was performed by following a model reaction, enzymatic hydrolysis of maltose, catalyzed by amyloglucosidase enzyme. The effects of substrate (maltose) and product (glucose) concentration on enzymatic reaction rate were studied. Data obtained from seven time courses were used for training of the ANN and another set of data obtained from eight time courses were used for testing of the trained network. This network was designed as a feed forward neural network with three neurons in the input layer, four neurons in the hidden layer and one neuron in the output layer. The network was trained till the mean square value between the targets and the outputs obtained was 1 × 10 −4 . The enzymatic reaction rate for the defined maltose and glucose concentration was estimated using the trained network. The regression coefficient of determination ( R 2 ) showed a good correlation between estimated and experimental data sets for both train (0.992) and test data sets (0.965). In further part of the study, estimated data by the ANN was used in a numerical solution of batch reactor modeling equation to obtain time courses data. There are high correlations between experimental and estimated time course curves and that was another proof of the high performance of ANN for estimation of enzyme-catalyzed reaction rate.