Simultaneous extraction of polysaccharides and polyphenols from blackcurrant fruits: Comparison between response surface methodology and artificial neural networks

A procedure for simultaneous extraction of polysaccharides and polyphenols from blackcurrant fruits by microwave-assisted aqueous two-phase system (MA-ATPS) was studied. Ethanol/(NH4)2SO4 was determined as the most suitable ATPS including 18.99 % (NH4)2SO4 and 31.65 % of absolute ethanol. Apart from response surface methodology (RSM), artificial neural network linked genetic algorithms and grey relational analysis (ANN-GA-GRA) was also employed to optimize the extraction process of polysaccharides and polyphenols from blackcurrant fruits. GA was used to give assistance to the parameters optimization of ANN model and GRA was subsequently utilized to optimize the variables. ANN model displayed relatively higher coefficient of determination (R2, 0.9805) and lower absolute average deviation (AAD, 0.6648) than those of RSM (R2 = 0.9783, AAD = 0.8505), indicating ANN possessed more excellent prediction and estimation capabilities. Under the optimum conditions (liquid-to-material ratio, extraction time, extraction temperature and microwave power predicted by RSM were 60.01 mL/g, 9.96 min, 59.57 °C, and 592.98 W, while predicted by ANN were 56.10 mL/g, 9.78 min, 60.92 °C, and 640.78 W), the yields of polysaccharides, anthocyanins, polyphenols, and DPPH radical scavenging activities of two phases given by ANN (46.75 %, 6.75 mg/g, 53.51 mg/g, 95.17 %, and 97.05 %) were higher than those deriving from RSM (43.48 %, 6.62 mg/g, 52.48 mg/g, 93.72 %, and 96.65 %). In addition, gas chromatography (GC) analysis confirmed that the polysaccharides in bottom phase were composed of six monosaccharides, and thirteen phenolic compounds were determined in the extract of top phase by high-performance liquid chromatography-electrospray ionization-mass spectrometry (HPLC-ESI-MS). Therefore, microwave-assisted ethanol/(NH4)2SO4 extraction may be a promising technique to simultaneously prepare multiple active compounds from blackcurrant, and ANN-GA-GRA was an effective mathematical model to predict and optimize this extraction process.