Main content area

Prediction of hyoscyamine content in Datura stramonium L. hairy roots using different modeling approaches: Response Surface Methodology (RSM), Artificial Neural Network (ANN) and Kriging

Amdoun, Ryad, Benyoussef, El-Hadi, Benamghar, Ahcene, Khelifi, Lakhdar
Biochemical engineering journal 2019 v.144 pp. 8-17
Datura stramonium, atropine, exposure duration, kriging, neural networks, prediction, response surface methodology, roots, salicylic acid
The comparison of different modelling approaches Response Surface Methodology (RSM), Artificial Neural Network (ANN) and Kriging, let to develop a novel ANN-RSM combined approach has been applied to accurately predict the variation of hyoscyamine content and especially to statistically explore the relationship between the response and the factors involved in the experiment of this study. The ANN, the Kriging and the polynomial models of 3 and 4 degrees model the complex elicitation response of Datura stramonium hairy roots elicited by salicylic acid. The ANN is the most accurate prediction approach followed by Kriging and RSM (polynomial models of 3 and 4 degrees). Unlike the ANN and the Kriging, the RSM has the advantage of statistically discriminating individuals effects of studied factors (exposure time, salicylic acid concentration) and their interaction on the response to the elicitation. So the approach, based on the combination of advantages of both ANN and RSM consisting in the application of the RSM with three 3² factorial designs to the responses accurately predicted by the ANN. The fit of quadratic models allowed to statistically exploring in optimal regions the factors influences and their interactions on the hyoscyamine content of hairy roots.