Prediction of Lactoferrin Concentrations in Mature Milk From Japanese Holstein Cows Using Milk Infrared Spectral Data and Its Practical Applicability

ABSTRACT

Lactoferrin is an important bioactive protein in milk associated with immune function in mammary glands. However, cost‐effective measurement of lactoferrin concentration in mature milk remains challenging, thus highlighting the potential of Fourier transform infrared (FTIR) spectroscopy as a predictive tool. To practically implement FTIR‐predicted lactoferrin concentrations, understanding both the methodological factors affecting prediction accuracy and the biological relevance of the predicted values is important. Therefore, this study aimed to identify the factors affecting the predictive accuracy of lactoferrin concentration in mature milk from Japanese Holstein cows using FTIR spectral data and evaluate the practical applicability of the predicted values. A total of 1624 milk samples were analyzed. Prediction models were developed using different spectral derivative preprocessing methods, spectral wavenumber ranges, and regression models. Predictive accuracies were generally comparable among the regression models, with Bayesian regularized neural network regression achieving the highest coefficient of determination ( R ²  = 0.57) under nonderivative preprocessing and spectral regions associated with milk quality traits. Environmental effects on the predicted lactoferrin concentrations showed patterns similar to those observed for the measured values, which suggests their potential applicability for herd‐level monitoring. These results indicate the promising potential of FTIR spectral information to predict lactoferrin concentration in mature milk.