Proceedings of the World Congress on Genetics Applied to Livestock Production, 11.225 Developing an indicator for body fat mobilisation using mid-infrared spectrometry of milk samples in dairy cows T. Mehtiö 1 , P. Mäntysaari 1 , T. Kokkonen 2 , S. Kajava 3 , T. Latomäki 4 , L. Nyholm 4 , C. Grelet 5 , T. Pitkänen 1 , E. A. Mäntysaari 1 & M.H. Lidauer 1 1 Natural Resources Institute Finland (Luke), Humppilantie 14, 31600 Jokioinen, Finland terhi.mehtio@luke.fi (Corresponding Author) 2 University of Helsinki, P.O. Box 28, 00014 Helsinki, Finland 3 Natural Resources Institute Finland (Luke), Halolantie 31 A, 71750 Maaninka, Finland 4 Valio Ltd., P.O. Box 10, 00039 Valio, Finland 5 Centre Wallon de Recherche Agronomiques, 24 Chaussée de Namur, 5030 Gembloux, Belgium Summary High energy requirement in the initiation of lactation may force dairy cows to mobilize energy from body tissue which leads to negative energy balance (EB). Negative EB can predispose cows to various health and fertility problems, and therefore should be considered in dairy cattle breeding programs especially if feed efficiency traits are included in breeding objectives. However, measuring of EB is difficult and estimates are imprecise. When a cow is in negative EB and mobilizing its fat reserves, concentrations of milk fatty acids and blood non-esterified fatty acids (NEFA) change. Therefore, it is possible to use these changes as biomarkers for energy status. Mid-infrared spectrometry (MIR) is a routinely used tool for milk samples and many milk fatty acids can be predicted with high accuracy by MIR. Our objective was to assess the capability of MIR to predict blood plasma NEFA concentration from milk samples. Milk and corresponding blood samples were collected from Nordic Red dairy cows in three research farms between 2013 and 2016. There were altogether 1585 milk spectral readings and 809 NEFA records from 141 cows in the data set. Partial least squares regression was used to predict NEFA and EB from MIR spectral data. The coefficient of determination of cross validation (R 2 cv) for NEFA was 0.64 when leave-one-out cross validation was used for the whole data set. Higher R 2 cv values were found when predicting blood plasma NEFA concentration from evening milk samples (0.67), probably because fatty acid concentrations in milk vary slower than NEFA concentration in blood. The robustness of the developed prediction equation was inspected by calibrating the equations with records from the cows from two research herds and then predicting NEFA for cows from the other herd, so that there were differences in the environment of the calibration data (130 NEFA obs) and test data (647 NEFA obs.) sets. Nevertheless, R 2 in the test data was 0.58 and RMSE 0.19 mmol/l, which indicates that the model is robust. Keywords: mid-infrared spectroscopy, non-esterified fatty acids, energy balance