Extraction of Citrullus colocynthis L. seed oil by supercritical carbon dioxide process using response surface methodology (RSM) and artificial neural network (ANN) approaches

In this research work, response surface methodology and artificial neural network were applied to Citrullus colocynthis seed oil extracted to optimize its oil yield (%) and antioxidant activities (IC50) by supercritical−CO2 extraction process and then compared. A Box Behnken design, with three-levels and five independent variables (pressure, extraction temperature, ethanol concentration, flow rate of CO2, and particle size), was also investigated. Besides, a backpropagation training-ANN with five inputs, one hidden layer (6 neurons) and one output, was selected as the best architectural topology of each tested response. Furthermore, the obtained data have revealed that the extraction pressure represents the most significant factor in oil yield (Y1) and the antioxidant activities (IC50) (Y2). Therefore, the optimum conditions, obtained by the model of artificial neural network, were as follow: pressure 275 bar, extraction temperature 72.4 °C, ethanol concentration 8.50 %, flow rate of supercritical−CO2 7.4 g/min, and particle size of 0.55 mm. Under these ultimate conditions, the oil yield (Y1) and the antioxidant activities (Y2) were 38.12 % and 0.17 mg/mL, respectively, which accords well with the values predicted by the ANN-model. Thus, although response surface methodology is likely to demonstrate the interaction effects between the independent variables of SC−CO2 extraction process, artificial neural network can reliably boost the SC−CO2 process with better estimative capabilities. Generally, the most elevated levels of bioactive compounds and the strongest antioxidant activities of Citrullus colocynthis SC−CO2 seed oils suggest that it could be economically used as appreciated natural products for chemical, cosmetic and pharmaceutical applications.