Mid-infrared milk screening as a phenotyping tool for feed efficiency in dairy cattle

Feed efficiency (FE) is one of the most essential traits in dairy cattle, primarily due to the high cost of feed, which constitutes a significant portion of dairy herd expenses. Unfortunately, assessing FE in individual cows requires precise measurement of feed consumption, a labour-intensive and expensive process that is impractical for group-fed cows on production farms. Efforts have been made to predict FE or, more precisely, dry matter intake (DMI), using predictors such as a body weight (BW), milk yield (MY), and milk composition. Recently, Fourier transform mid-infrared (FT-MIR) spectroscopy has been proposed as a tool to enhance the accuracy of DMI prediction. This paper reviews the application of FT-MIR milk spectroscopy for deriving FE phenotype in dairy cattle. FT-MIR is a reliable and widely used method for routine analysis of milk components. In FE phenotyping, predictive equations often incorporate FT-MIR alongside other traits such as BW, MY, milk composition, herd, breed, days in milk, and pregnancy. The most commonly used mathematical approaches are partial least squares (PLS) regression and artificial neural networks (ANN). Prediction accuracy varies across studies, depending on the mathematical method and model employed. Predictions based solely on FT-MIR data have demonstrated moderate accuracy (coefficient of determination), ranging from 0.19 to 0.40. However, integrating all data sources including MY, milk composition, FT-MIR, and near-infrared reflectance spectroscopy (NIR) is crucial and results in higher accuracy, with reported values ranging from 0.03 to 0.81.