Authentication of P.G.I. Gragnano pasta by near infrared (NIR) spectroscopy and chemometrics

• Published in: Microchemical journal Vol. 152; p. 104339
• Main Authors: Firmani, Patrizia, La Piscopia, Giuseppe, Bucci, Remo, Marini, Federico, Biancolillo, Alessandra
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.01.2020
• Subjects: Classification; Durum wheat pasta; Gragnano; Near infrared (NIR) spectroscopy; Partial least squares-discriminant analysis (PLS-DA); Soft independent modeling of class analogies (SIMCA); Durum wheat pasta; Near infrared (NIR) spectroscopy; Gragnano; Partial least squares-discriminant analysis (PLS-DA); Classification; Soft independent modeling of class analogies (SIMCA)
• ISSN: 0026-265X; 1095-9149
• DOI: 10.1016/j.microc.2019.104339

•Gragnano Pasta is a PGI Italian durum pasta.•949 samples of pasta were analyzed by NIR coupled with PLS-DA or SIMCA.•Classifiers were used to distinguish Gragnano and Non-Gragnano pasta.•PLS-DA misclassified 1 sample over 200 of test (external validation).•SIMCA led to 96.57% sensitivity and 100% specificity (external validation). Pasta is a typical Italian food item obtained by durum wheat semolina/flour well-known and widely consumed all over the world. Since 2013, Gragnano Pasta, a typical aliment produced in a specific area in the South of Italy, has been awarded with the P.G.I. mark, remarking the high value of this product. Due to its peculiarity and its market value, it is important to characterize and authenticate the Gragnano Pasta. Considering this rationale, the present study aims at developing a non-destructive analytical methodology suitable for this goal. Consequently, the possibility of coupling Near Infrared spectroscopy (NIR) with two different classifiers has been tested. In particular, 949 samples of pasta were analysed, and then classified into categories Gragnano and non-Gragnano by Partial Least Squares Discriminant Analysis (PLS-DA) and Soft Independent Modeling of Class Analogies (SIMCA). In order to externally validate models, samples were divided into a training and a test set of 749 and 200 objects, respectively. Both approaches provided excellent results; PLS-DA correctly classified all the Gragnano samples (and it misclassified only 1 object belonging to the other category), while SIMCA analysis (modelling Class Gragnano) led to 96.57% sensitivity and 100% specificity.