Prediction of protein composition of individual cow milk using mid-infrared spectroscopy

Abstract

This study investigated the application of mid-infrared spectroscopy for the prediction of protein composition in individual milk samples (n=1,336) of Simmental cows. Protein fractions were quantified by RP-HPLC and MIR data were recorded over the spectral range from 4,000 to 900 cm-¹. Models were developed by partial least squares regression using untreated spectra. The most successful predictions were for protein, casein, α_S1-casein, whey protein, and β-lactoglobulin contents. The models could discriminate between high and low values of protein composition (R²=0.50 to 0.58). The root mean square errors of cross-validation were 3.11g/l for protein (range 39.91g/l), 2.76g/l for casein (range 35.16g/l), 1.07 g/l for α_S1-casein (range 12.82g/l), 0.51 g/l for whey protein (range 4.97g/l), and 0.43 for β-lactoglobulin (range 4.37). Application of MIR spectroscopy is possible for the routinely assessment of protein, casein, α_S1-casein, whey protein, and β-lactoglobulin and its implementation might be in future a tool for impro ving protein composition of bovine milk through breeding programs.

Keywords:

• Milk
• Protein composition
• Mid-infrared spectroscopy
• Chemometrics