Constitution of an international dataset on blood biomarkers in dairy cows: a preliminary study to develop milk MIR models

Blood composition remains the gold standard to monitor and detect various health disorders of dairy cows. Estimating blood components through non-invasive methods would enable to scale up the measures in terms of cow number and time frequency, while aligning with societal requirements. This work begins with the constitution of a large dataset from multiple organizations across 12 countries. The first objective was to conduct an explanatory study, to better understand the variability of blood biomarkers regarding animal characteristics, sampling protocols and their relationship with other phenotypes of interest. The second objective was to improve on the large variability to develop robust models based on milk MIR spectra. Data merging resulted in a dataset of approximately 10,000 individual records of blood reference values and associated milk spectra. The majority of records were associated with blood BHB and NEFA, and fewer records with glucose, IGF-I, fructosamine, cholesterol, urea, progesterone, calcium and phosphorus. This preliminary work will facilitate a better understanding of the sources of variability in biomarkers, to highlight optimal modelling methodologies among linear and non-linear algorithms, and to estimate the capacities of milk MIR spectra to provide information on those traits under routine conditions,.