Potential of milk mid-infrared spectra to predict nitrogen use efficiency of individual dairy cows in early lactation

摘要

Improving nitrogen use efficiency (NUE) at boththe individual cow and the herd level has become akey target in dairy production systems, for both environmentaland economic reasons. Cost-effective andlarge-scale phenotyping methods are required to improveNUE through genetic selection and by feedingand management strategies. The aim of this study wasto evaluate the possibility of using mid-infrared (MIR)spectra of milk to predict individual dairy cow NUEduring early lactation. Data were collected from 129Holstein cows, from calving until 50 d in milk, in 3research herds (Denmark, Ireland, and the UK). In 2 ofthe herds, diets were designed to challenge cows metabolically,whereas a diet reflecting local managementpractices was offered in the third herd. Nitrogen intake(kg/d) and nitrogen excreted in milk (kg/d) were calculateddaily. Nitrogen use efficiency was calculated asthe ratio between nitrogen in milk and nitrogen intake,and expressed as a percentage. Individual daily valuesfor NUE ranged from 9.7 to 81.7%, with an averageof 36.9% and standard deviation of 10.4%. Milk MIRspectra were recorded twice weekly and were standardizedinto a common format to avoid bias between apparatusor sampling periods. Regression models predictingNUE using milk MIR spectra were developed on 1,034observations using partial least squares or supportvector machines regression methods. The models werethen evaluated through (1) a cross-validation using 10subsets, (2) a cow validation excluding 25% of the cowsto be used as a validation set, and (3) a diet validationexcluding each of the diets one by one to be usedas validation sets. The best statistical performanceswere obtained when using the support vector machinesmethod. Inclusion of milk yield and lactation numberas predictors, in combination with the spectra, alsoimproved the calibration. In cross-validation, the bestmodel predicted NUE with a coefficient of determinationof cross-validation of 0.74 and a relative error of14%, which is suitable to discriminate between low- andhigh-NUE cows. When performing the cow validation,the relative error remained at 14%, and during the dietvalidation the relative error ranged from 12 to 34%.In the diet validation, the models showed a lack of robustness,demonstrating difficulties in predicting NUEfor diets and for samples that were not represented inthe calibration data set. Hence, a need exists to inte-grate more data in the models to cover a maximumof variability regarding breeds, diets, lactation stages,management practices, seasons, MIR instruments, andgeographic regions. Although the model needs to bevalidated and improved for use in routine conditions,these preliminary results showed that it was possible toobtain information on NUE through milk MIR spectra.This could potentially allow large-scale predictions toaid both further genetic and genomic studies, and thedevelopment of farm management tools.