The Fourier-transform infrared spectroscopy (FTIR) is one of the most developed and implemented tool for the analysis of milk chemical compounds. Besides this, FTIR can also be used to determine the fingerprint of milk for authentication purposes to certify the area of origin or the farming system in which the milk is produced. The present study, carried out within the INTAQT EU project, aimed at assessing the effectiveness of FTIR applied to bulk milk samples in discriminating dairy herds of Parmigiano Reggiano Consortium (PRC) for their structural and management characteristics. Dairy farm information included altitude zone (AZ), herd size (HS), housing type (HT), dairy cows genetic type (GT), use of total mixed rations (MR), and proportion of concentrate inclusion in the cow diets (CONC). This database was merged with milk data obtained from the official milk recording system along with FTIR spectral data of bulk milk, stored by the Breeders Association of Emilia Romagna Region lab (ARAER, Reggio Emilia, Italy) from January to August 2022. Overall data set comprised 4,610 bulk milk FTIR spectra from 940 farms, with a mean of 4.9 (±1.1) observations per farm respectively. Each spectrum contained absorbance values at 1,060 different wavenumbers (5,000 to 930 × cm-1). Quality control of spectral data involved centering and scaling, removing spectral samples using a Mahalanobis distance greater than 3 SD, and removing the water region. A Partial Least Squares Discriminant Analysis (PLS-DA) and Linear Discriminant Analysis (LDA) were fitted to estimate the probability of each observation (i.e. farm) belonging to a specific group (e.g. AZ, HS, etc.). The 60% of the data was used as training set and 40% as testing set. Finally, a prediction of the testing set was performed with both methods. PLS-DA gained an accuracy of 85, 74, 86, 82 and 95% for classifying the proportion of CONC, AZ, HT, MR and GT respectively. In the case of LDA the accuracy values obtained were 63, 73, 86, 82 and 97% respectively. These results suggest the potential of FTIR to determine the fingerprint of milk for authentication purposes in the PRC area.
Potential of milk infrared spectroscopy to discriminate farm characteristics: the INTAQT project
Marco Aurelio Ramirez Mauricio;Diana Giannuzzi;Luigi Gallo;Marco Berton;Alessio Cecchinato;Enrico Sturaro
2023
Abstract
The Fourier-transform infrared spectroscopy (FTIR) is one of the most developed and implemented tool for the analysis of milk chemical compounds. Besides this, FTIR can also be used to determine the fingerprint of milk for authentication purposes to certify the area of origin or the farming system in which the milk is produced. The present study, carried out within the INTAQT EU project, aimed at assessing the effectiveness of FTIR applied to bulk milk samples in discriminating dairy herds of Parmigiano Reggiano Consortium (PRC) for their structural and management characteristics. Dairy farm information included altitude zone (AZ), herd size (HS), housing type (HT), dairy cows genetic type (GT), use of total mixed rations (MR), and proportion of concentrate inclusion in the cow diets (CONC). This database was merged with milk data obtained from the official milk recording system along with FTIR spectral data of bulk milk, stored by the Breeders Association of Emilia Romagna Region lab (ARAER, Reggio Emilia, Italy) from January to August 2022. Overall data set comprised 4,610 bulk milk FTIR spectra from 940 farms, with a mean of 4.9 (±1.1) observations per farm respectively. Each spectrum contained absorbance values at 1,060 different wavenumbers (5,000 to 930 × cm-1). Quality control of spectral data involved centering and scaling, removing spectral samples using a Mahalanobis distance greater than 3 SD, and removing the water region. A Partial Least Squares Discriminant Analysis (PLS-DA) and Linear Discriminant Analysis (LDA) were fitted to estimate the probability of each observation (i.e. farm) belonging to a specific group (e.g. AZ, HS, etc.). The 60% of the data was used as training set and 40% as testing set. Finally, a prediction of the testing set was performed with both methods. PLS-DA gained an accuracy of 85, 74, 86, 82 and 95% for classifying the proportion of CONC, AZ, HT, MR and GT respectively. In the case of LDA the accuracy values obtained were 63, 73, 86, 82 and 97% respectively. These results suggest the potential of FTIR to determine the fingerprint of milk for authentication purposes in the PRC area.
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
