Potential of High-Performance Liquid Chromatography with Ultraviolet Detection (HPLC-UV) Fingerprints to Assess the Geographical Production Origin and Authenticity of Honey
Honey is a widely appreciated and consumed natural product which is highly susceptible to fraudulent practices involving different sample attributes such as the botanical species or the geographical production regions, as well as possible adulterations. In the present work, the potential of non-targ...
| Autores: | , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Universidad de Barcelona |
| Repositorio: | Dipòsit Digital de la UB |
| OAI Identifier: | oai:diposit.ub.edu:2445/217969 |
| Acceso en línea: | https://hdl.handle.net/2445/217969 |
| Access Level: | acceso embargado |
| Palabra clave: | Mel d'abelles Cromatografia de líquids d'alta resolució Quimiometria Honey High performance liquid chromatography Chemometrics |
| Sumario: | Honey is a widely appreciated and consumed natural product which is highly susceptible to fraudulent practices involving different sample attributes such as the botanical species or the geographical production regions, as well as possible adulterations. In the present work, the potential of non-targeted HPLC-UV fingerprints as honey chemical descriptors to assess their geographical origin authentication involving a high number of samples belonging to nine different countries (and 4 continents) was evaluated by partial least squares-discriminant analysis (PLS-DA). Accurate discrimination between Spanish and Italian samples independently of the botanical varieties involved (multifloral, rosemary, and eucalyptus) was accomplished, as well as for the botanical species discrimination when considering each country independently. The best classification performance for 157 honey samples produced in 9 countries was accomplished when HPLC-UV fingerprints were submitted to a classification decision tree performed by consecutive PLS-DA models built using hierarchical model builder (HMB), with sensitivity and specificity values (for calibration and cross-validation) higher than 87.5 and 78.6%, respectively, and with classification errors below 17.0%. Prediction capabilities improved for samples belonging to New Zealand, Costa Rica, The Netherlands, and China, with classification errors below 8.3%, while it worsened for the other sample groups (classification errors in the range 17.4-27.4% for the samples belonging to Spain, Italy, France, and Serbia). Japanese samples showed the worse prediction errors (37.5%) as the “unknown” samples used were mostly misclassified as Chinese samples. |
|---|