Artificial Neural Network Modeling To Predict And Optimize Phenolic Acids Production From Callus Culture of Lactuca Undulata

Abstract

The present study aims to model and optimize phenolic acids productions from Lactuca undulate root and leaf-derived callus using the feed-forward Artificial Neural Network (ANN) model. For this purpose, the effect of different concentrations (0, 0.1, 0.5, 1, and 2 mg/l) of Kin in combinations with or without 2,4-D and/or NAA was investigated on callus induction and phenolic acids production. A multi-layer perceptron ANN was applied to correlate the output parameters (cichoric acid, chlorogenic acid and caffeic acid contents) to input (Kin, 2,4-D and NAA) training parameters. A single hidden layer with 5, 10, 15, 20, 25, 30, 35 and 40 neurons was used to optimize ANN architecture. Sum squared error (SSE), Relative Error (RE) and correlation factor (R2) were applied to identify the performance of ANN models. According to the obtained data, the feed-forward neural network with tangent-sigmoid (3-30-1), tangent-tangent (3-15-1) and tangent-tangent (3-35-1) activation function was found as the best model to predict cichoric acid, chlorogenic acid and caffeic acid production from leaf-derived callus, respectively. Meanwhile, ANN with activation function of tangent-tangent (3-20-1), tangent-tangent (3-25-1) and sigmoid-sigmoid (3-20-1) were the most effective models to predict the amount of cichoric acid, chlorogenic acid and caffeic acid from root-derived callus, respectively. In the current study, there was a strong correlation between experimental and predicted data. These results demonstrated that the selected ANN model could predict the effects of plant growth regulators on phenolic acids production using callus culture method.