Green extraction and optimization of bioactive compounds from Solanum torvum Swartz. using ultrasound-aided solvent extraction method through RSM, ANFIS and machine learning algorithm

Abstract

Biologically important phytocompounds based on total polyphenolic, flavonoid contents and their free radical scavenging potential were completely harvested from Solanum torvum Sw. through ultrasound-aided solvent extraction (UASE) method. The face-centered central composite design (FCCD) in response surface methodology (RSM) was applied to optimize the effective extraction parameters, and adaptive neuro-fuzzy inference system (ANFIS), as well as machine learning (ML) algorithm models were used to validate the optimized extraction parameters for the highest yield of bioactive phytocompounds. The effect of five extraction parameters, namely methanol concentration (X1: 55–65% in v/v with water), ultrasound-intensity (X2: 50–60 W cm−2), temperature (X3: 30–45 °C), time (X4: 15–25 min) and particle size (X5: 0.5–1.5 mm) at five levels (very low (−2.37), low (−1), medium (0), high (+1) and very high (+2.37)) were studied for optimized yields of total polyphenolic (TP; (y1)), total flavonoid contents (TF; (y2)) and their free radical scavenging abilities (%DPPH*sc (y3), %ABTS*sc (y4), and %H2O2*sc (y5)). The optimal condition was attained at X1 = 65%, X2 = 60 W cm−2, X3 = 45 °C, X4 = 15 min, and X5 = 0.5 mm; under this condition, the highest yields of y1 = 830.677 mg gallic acid equivalent (GAE)/g, y2 = 535.38 mg rutin equivalents (RE)/g, y3 = 75.22%, y4 = 84.12%, and y5 = 85.02% was obtained. Most of the experimental values were well matched with the predicted responses of the RSM model. Further, ANFIS and random forest-machine learning algorithm were utilized to compare and validate the optimized extraction parameters. Also, GC-MS and LC-MS were carried out to investigate the presence of various bioactive phytocompounds in the optimized extract.