Abstract
Abstract The official method for the determination of antioxidant activity in beverages is 2,2-diphenyl-1-picrylhydrazyl (DPPH) assay, which requires toxic reagents, is laborious, and produces waste. The aim of this work was to develop a more eco‑friendly method for the prediction of antioxidant activity in wine brandy and sweet wine using synchronous fluorescence spectra (SFS). In scanning of bulk and diluted samples, the excitation wavelength was varied from 250 to 500 nm and the wavelength interval was ranged from 20 to 100 nm. Partial least squares (PLS) regression was done on individual SFS, on unfolded SFS and on variables selected by the variable importance in the projection (VIP) algorithm, while the DPPH assay was the reference method. VIP-PLS modeling of the SFS of diluted samples led to better performance characteristics of the regression models. The best VIP-PLS model for wine brandy with relative predictive deviation (RPD) of 3.9 was based on 62 variables (the wavelength interval from 80 to 100 nm and the excitation wavelength from 290 to 320 nm). The best VIP-PLS model for sweet wine with RPD of 4.2 was calculated on 108 variables (the wavelength interval from 60 to 100 nm and the excitation wavelength from 260 to 290 nm). RPD values above 3.5 indicated very good prediction accuracy obtained by VIP-PLS models. Analytical GREEnness (AGREE) score 0.74 confirmed a high level of greenness of the proposed method.
Funder
Slovak Technical University
Publisher
Springer Science and Business Media LLC