Analytical Methods Varietal discrimination of Chilean wines by direct injection mass spectrometry analysis combined with multivariate statistics Evelyn Villagra a , Leonardo S. Santos b , Boniek Gontijo Vaz c , Marcos N. Eberlin c , V. Felipe Laurie a,⇑ a School of Agricultural Sciences, Universidad de Talca, 2 Norte 685, Talca, Chile b Asymmetric Synthesis Laboratory, Chemistry Institute of Natural Resources, Universidad de Talca, 2 Norte 685, Talca, Chile c Thomson Mass Spectrometry Laboratory, Chemistry Institute, University of Campinas, UNICAMP, 13083-970 Campinas, SP, Brazil article info Article history: Received 1 June 2011 Received in revised form 13 July 2011 Accepted 26 August 2011 Available online 10 September 2011 Keywords: Wine Discrimination Mass spectrometry Multivariate statistics Fingerprinting abstract A simple, direct injection, electrospray ionization Fourier transform mass spectrometry (ESI FT–MS) method, in combination with multivariate statistics, was used for the characterization and sorting of Chil- ean wines. 47 commercial red wines labelled as Cabernet Sauvignon, Carménère, Syrah, and Pinot noir, and 25 white wines of the varieties Chardonnay and Sauvignon blanc were diluted, directly infused into the mass spectrometer, and analyzed in negative ion mode. The signature ions used for statistical anal- yses were manually filtered out from the signals with m/z ratios over 10%. The results of principal com- ponent analysis allowed a good sorting of white wines, but not so in the case of reds. The main three principal components explained 96.82% and 85.65% of the variance for white and red wines, respectively. Instead, linear discriminant analysis, allowed the correct discrimination of 100.00% of white and 95.74% of red samples. The validation of these results using the leave-one-out cross-validation method gave lower percentages of correct classification (76.00% and 61.70% of white and red samples respectively), suggesting that some of the wine samples analyzed might have been blends of more than one variety. Consequently, ESI FT–MS direct injection analysis of wines can be used for sample discrimination, but requires a stronger mathematical model built with spectral information of pure and blended samples before improving the percentages of classification. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction Researchers have developed and optimized a series of analytical methods that combined with chemometric analyses, allow for the accurate assessments of the composition, origin, and possible adulteration of food products. A few examples of studies published in this regard include: the assessment of the geographical origin of honey (Jun, Michael, Soojin, Mel, & Qing, 2010), tea (Moreda-Pineiro, Fisher, & Hill, 2003), and fresh fruits (Perez, Smith, & Anderson, 2006), organic composition of coffee to determinate ripeness (Amo- rim et al., 2009), cachaça with different types of ageing (de Souza et al., 2007) and manufacturing processes (de Souza et al., 2009). These studies have suggested some of the key components (i.e. organic or inorganic) of these food matrices that may be related to a given geographical area, type of preparation, variety, etc. In wines, there have been also a few studies of this type (Saurina, 2010), both for the evaluation of provenance (Bellomarino, Conlan, Parker, Barnett, & Adams, 2009; Cynkar, Dambergs, Smith, & Cozzolino, 2010; Forina, Oliveri, Jäger, Römisch, & Smeyers-Verbeke, 2009; Marini, Bucci, Magri, & Magri, 2006) and composition (Chambery, del Monaco, Di Maro, & Parente, 2009; Louw et al., 2009; Pena-Neira, Hernandez, Garcia-Vallejo, Estrella, & Suarez, 2000). These studies have shown that it is possible to determine a pattern indicating the source of the products, and some of their qual- ity features, based on the analysis of their organic and inorganic con- stituents (Saurina, 2010); thus taking advantage of the various changes occurred in the chemical composition of grapes and wines during production and processing stages. Some of the more reliable and widely used techniques to determine the organic composition of wines are those coupling chromatography with mass spectrometry; particularly, HPLC–MS (Lamuela-Raventos & Waterhouse, 1994) and GC–MS (Ortega, López, Cacho, & Ferreira, 2001). However, the use of these techniques might be time consuming, primarily due to sample pre-treatment and the time required for the subsequent chromato- graphic separation (De Beer et al., 2004). In practice, it would be desirable that the methods of analyses used did not oblige a compli- cated sample pre-treatment, and that the time required for chemical analysis and chemometrics were as short as possible. Lately, the use of instruments that allow obtaining chemical information from complex samples by direct injection, has permitted a notable reduc- tion of the time spent during the chemical analyses of organic com- pounds. Some examples of the prior are: the identification of the organic composition of red wines by measuring polyphenols, organ- ic acids, alcohols, and nitrogen-based compounds, by MALDI–MS 0308-8146/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodchem.2011.08.078 ⇑ Corresponding author. Tel.: +56 71200214; fax: +56 71200212. E-mail address: flaurie@utalca.cl (V. Felipe Laurie). Food Chemistry 131 (2012) 692–697 Contents lists available at SciVerse ScienceDirect Food Chemistry journal homepage: www.elsevier.com/locate/foodchem