Optimization of ultrasound phenolic extraction from Erica multiflora leaves using response surface methodology and artificial neural networks

Abstract

This study aimed to develop and optimize an efficient ultrasound-assisted extraction method for phenolic compounds from Erica multiflora leaves, with the primary objectives of preserving extract quality, minimizing solvent usage, and standardizing extraction conditions. A total of 15 extractions were performed, varying solvent concentration, extraction time, and ultrasonic amplitude, using a Box-Behnken design. The effects of the extraction parameters on phenolic yield were modeled using response surface methodology (RSM) and artificial neural networks (ANN). Both models demonstrated strong predictive capabilities for all measured responses. Among the extraction parameters, amplitude and time were identified as key factors influencing phenolic yield. The optimized conditions 50.79 % solvent concentration, 60 seconds extraction time, and 86.38 % ultrasonic amplitude resulted in a TPC of 106.42 ± 3.79 mg gallic acid equivalents (GAE) per gram of dry weight (DW), a TFC of 14.37 ± 0.51 mg quercetin equivalents (QE) per gram DW, and a DPPH antioxidant capacity of 3.11 ± 0.85 mg ascorbic acid equivalents (AE) per gram DW. The experimental results closely matched the model predictions, validating the optimization process. The methanolic extracts of Erica multiflora demonstrated significant potential for use in pharmaceuticals, nutraceuticals, and functional foods.