Determination of total phenolic content and selected phenolic compounds in sweet wines by fluorescence spectroscopy and multivariate calibration

The aim of this work was to develop a simple, fast and reagent-free method for the prediction of total phenolic content (TPC) and selected phenolic compounds in Slovak botrytized Tokaj wines based on spectral data and partial least squares (PLS) regression. The significant variables for the prediction using PLS were obtained by evaluating the variable importance in projection (VIP). Folin-Ciocalteu and HPLC, respectively, were used as reference methods for the determination of TPC and selected phenolic compounds. Using UV-vis and excitation-emission matrix (EEM) fluorescence spectra recorded on the bulk and diluted samples, the calibration models for TPC were developed and compared. The best PLS model with relative predictive deviation (RPD) of 5.8, coefficient of determination (R2) of 0.972 and root mean squares error (RMSE) of 20.2 mg/L was based on the variables selected from unfolded EEM fluorescence spectra of diluted samples. Using unfolded EEM fluorescence spectra recorded on the diluted samples and VIP, the PLS calibration models for gallic, protocatechic, caffeic and p-coumaric acids and (+)-catechin were obtained with RPD > 4.0 and R2 > 0.9. The RMSE values were 0.7, 0.5, 0.3, 0.2 and 1.0 mg/L for gallic, protocatechic, caffeic and p-coumaric acids and (+)-catechin, respectively.

Automated summary

The aim of this study was to develop a simple, fast, and reagent-free method for predicting the total phenolic content and selected phenolic compounds in Slovak botrytized Tokaj wines using spectral data and partial least squares regression. The best predictive model achieved an RPD of 5.8, R2 of 0.972, and RMSE of 20.2 mg/L for TPC. The PLS calibration models for gallic, protocatechic, caffeic, p-coumaric acids, and (+)-catechin had high prediction accuracy with RPD > 4.0 and R2 > 0.9.