An exploratory chemometric study of 1H NMR spectra of table wines

In this paper an exploratory study of 40 table wines by proton nuclear magnetic resonance spectroscopy and chemometric region-selection methods is presented. Several components of wine have been identified and quantified. It is demonstrated how signal alignment procedures were utilized to compensate for pH effects in the NMR spectra prior to the chemometric data modeling. The analysis included region selection by interval partial least squares for regression to reference data-obtained from infrared spectroscopy. Accurate calibration models to the contents of ethanol, glycerol, lactic acid, methanol and malic acid were established. For the more general combined glucose/fructose infrared reference backwards interval partial least squares was introduced for optimal interval selection in calibration. The aim of the paper is to show how pre-processing and region-selection methods can assist in interpretation and quantification of chemical shift multiplets in H-1 NMR spectra of complex biological systems. The extension to NMR metabolomics is straightforward. Copyright (C) 2006 John Wiley & Sons, Ltd.