The aim of this study was to predict the geographic origin of lentils by using isotope ratio mass spectrometry (IRMS) in combination with chemometrics. Lentil samples from two origins, i.e. Italy and Canada, were analysed obtaining the stable isotope ratios of δ<sup>13</sup>C, δ<sup>15</sup>N, δ<sup>2</sup>H, δ<sup>18</sup>O, and δ<sup>34</sup>S. A comparison between median values (U-test) highlighted statistically significant differences (p < 0.05) for all isotopic parameters between the lentils produced in these two different geographic areas, except for δ<sup>15</sup>N. Applying principal component analysis, grouping of samples was observed on the basis of origin but with overlapping zones; consequently, two supervised discriminant techniques, i.e. partial least squares discriminant analysis and k-nearest neighbours algorithm were used. Both models showed good performances with external prediction abilities of about 93% demonstrating the suitability of the methods developed. Subsequently, isotopic determinations were also performed on the protein and starch fractions and the relevant results are reported.
Discrimination of geographical origin of lentils (Lens culinaris Medik.) using isotope ratio mass spectrometry combined with chemometrics
LONGOBARDI, FRANCESCO;CASIELLO, GRAZIA;CATUCCI, Lucia;AGOSTIANO, Angela
2015-01-01
Abstract
The aim of this study was to predict the geographic origin of lentils by using isotope ratio mass spectrometry (IRMS) in combination with chemometrics. Lentil samples from two origins, i.e. Italy and Canada, were analysed obtaining the stable isotope ratios of δ13C, δ15N, δ2H, δ18O, and δ34S. A comparison between median values (U-test) highlighted statistically significant differences (p < 0.05) for all isotopic parameters between the lentils produced in these two different geographic areas, except for δ15N. Applying principal component analysis, grouping of samples was observed on the basis of origin but with overlapping zones; consequently, two supervised discriminant techniques, i.e. partial least squares discriminant analysis and k-nearest neighbours algorithm were used. Both models showed good performances with external prediction abilities of about 93% demonstrating the suitability of the methods developed. Subsequently, isotopic determinations were also performed on the protein and starch fractions and the relevant results are reported.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.