Artificial neural network optimisation of natural deep eutectic solvent extraction process of Danshen–Gegen and cytotoxicity study

Abstract

Natural deep eutectic solvents (NaDESs) are environmentally friendly and efficient for the componential extraction of traditional Chinese medicine compared to conventional organic solvents. In this study, NaDES was screened and employed to extract Danshen–Gegen (DG), and the extraction process was optimised by response surface methodology (RSM) and artificial neural networks (ANN) model. Besides, the in vitro security of extracts of DG were evaluated in PC12 cells. As a result, Betaine-Urea (Bet-Ur) was screened as extraction solvent and ANN model was more accurate than RSM model in optimising the extraction parameter. The extraction process optimised by ANN was as follows: 70% NaDES concentration, 80 mg/mL solid to liquid ratio, 67 °C ultrasonic temperature, and 33 min of ultrasonic time. The comprehensive value of extraction yield was 0.7251 ± 0.84%. IC50 of Bet-Ur, NaDES DG extract and aqueous DG extract were 0.15%, 0.3% and 10% (v/v).