Assessment of infrared spectroscopy for the quantification of protein fractions in buffalo milk

Despite the subject of cow milk proteins have been extensively addressed, there still is a lack of studies investigating detailed protein profile in buffalo milk. The aim of this study was to qualitatively and quantitatively determine protein fractions in buffalo milk. An additional aim was to assess the effectiveness of mid-infrared spectroscopy to predict milk protein composition. Individual milk samples of Italian Mediterranean buffaloes (n = 401) were collected in 7 commercial herds located in the area of Mozzarella di Bufala Campana PDO (south of Italy). Traditional milk quality traits (fat, protein, casein and lactose, %), pH and urea content (mg/dL) and spectral data were obtained using MilkoScan FT6000 (Foss Electric A/S, Hillerød, Denmark). Detailed milk protein composition (αS1-, αS2-, β-, and κ-casein, α-lactalbumin and β-lactoglobulin) were determined by reversed-phase HPLC. Within sample, the concentration of each protein fraction was matched with the corresponding spectrum. Chemometric analyses were carried out using WinISI 4 software (Infrasoft International, Port Matilda, PA) applying a 5-fold cross-validation. Results showed moderate accuracies of the mid-infrared spectroscopy to predict different protein fractions, with a coefficient of determination in cross validation (R2) from 0.34 (β-lactoglobulin) to 0.60 (β-casein). In perspective, further research should be addressed to improve the accuracy of near-infrared spectroscopy models by increasing the number of samples in calibration, or considering alternative chemometric approaches.