Use of an artificial neural network in modeling yeast biomass and yield of β-glucan

Abstract

An artificial neural network (ANN) approach was used to model dry yeast cell mass ( Saccharomyces cerevisiae NCIM 3458) and its glucan contents. The dry cell mass and yield of glucan depend mainly on concentration of media components. The mathematical relationship between them was not well established. This relationship between the concentration of media components with yield of glucan and with dry cell mass is important from a process optimization and control point of view. In the present work an ANN model was developed, which incorporated the effect of the following media components: glucose, peptone, yeast extract, malt extract, Mn 2+ and Mg 2+ as input parameters and the yield glucan and dry biomass is obtained after optimizing model parameters. The predictive capacity of the trained network was tested using separate data set (testing set), which is not used for the training. The average quadratic error (AQE) for trained network was 0.0012 and 0.003 for glucan and biomass, respectively. The network predicts the yield of glucan within the range ±3.5% and biomass within the rage ±5.5% of the experimental values. The trained network showed comparable trends in change of yield of glucan with change in respective input parameters.