Prediction of proteins including free amino acids in bovine milk by mid-infrared spectroscopy

The aim of this study was to evaluate the effectiveness of mid-infrared spectroscopy (MIRS) in predicting individual milk proteins and free amino acids in bovine milk and estimating their correlations with milk processing characteristics. A total of 730 milk samples were collected from seven Irish research herds and represented cows from a range of breeds, parities and stages of lactation. Gold-standard methods were used to quantify protein fractions and free amino acids (FAA) of these samples. Separate prediction equations were developed for each trait using partial least squares regression and accuracy of prediction was assessed using both cross validation and external validation. The greatest coefficient of correlation in external validation obtained for protein fractions was for total casein (0.74), while weak to moderate prediction accuracies were observed for FAA and among these glycine had the greatest coefficient of correlation (0.75 in both cross validation and external validation). Near unity correlations existed between total casein and beta-casein irrespective of whether the traits were based on the gold-standard (0.92) or MIRS predictions (0.95). Pearson correlations between gold-standard protein fractions and gold-standard milk processing characteristics ranged from -0.48 (pH and protein) to 0.50 (total casein and curd firmness). The lactation profile of total FAA indicated that the greatest concentration of FAA in milk was during early and late lactation. Results from this study demonstrate that mid-infrared spectroscopy has the potential to predict protein fractions and some FAA in milk.