Torres Cobos, BertaQuintanilla-Casas, BeatrizRovira, MercèRomero, AgustíGuardiola Ibarz, FrancescVichi, S. (Stefania)Tres Oliver, Alba2025-03-172025-03-172023-12-260308-8146https://hdl.handle.net/2445/219758<p>This study compares two data processing techniques (fingerprinting and untargeted profiling) to authenticate</p><p>hazelnut cultivar and provenance based on its unsaponifiable fraction by GC–MS. PLS-DA classification models</p><p>were developed on a selected sample set (n = 176). As test cases, cultivar models were developed for “Tonda di</p><p>Giffoni” vs other cultivars, whereas provenance models were developed for three origins (Chile, Italy or Spain).</p><p>Both fingerprinting and untargeted profiling successfully classified hazelnuts by cultivar or provenance,</p><p>revealing the potential of the unsaponifiable fraction. External validation provided over 90 % correct classification,</p><p>with fingerprinting slightly outperforming. Analysing PLS-DA models’ regression coefficients and</p><p>tentatively identifying compounds corresponding to highly relevant variables showed consistent agreement in</p><p>key discriminant compounds across both approaches. However, fingerprinting in selected ion mode extracted</p><p>slightly more information from chromatographic data, including minor discriminant species. Conversely,</p><p>untargeted profiling acquired in full scan mode, provided pure spectra, facilitating chemical interpretability.</p>1 p.application/pdfengcc-by-nc-nd (c) Berta Torres-Cobos, et al., 2023http://creativecommons.org/licenses/by-nc-nd/4.0/AvellanesQuímica dels alimentsHazelnutsFood compositioninfo:eu-repo/semantics/article7455822025-03-17info:eu-repo/semantics/openAccess