Abstract #752

# 752
Variation in milk fat globule size in Canadian dairy cattle and its prediction using mid-infrared spectroscopy.
Allison Fleming*1, Astrid Koeck1, Flavio Schenkel1, Milena Corredig2,3, Mehdi Sargolzaei1,4, Bonnie Mallard5, R. Ayesha Ali6, Saranya Gunasegaram1, Filippo Miglior1,7, 1CGIL, University of Guelph, Guelph, ON, Canada, 2Gay Lea Foods, Mississauga ON, Canada, 3Dept of Food Science, University of Guelph, Guelph, ON, Canada, 4Semex, Guelph, ON, Canada, 5Dept of Pathobiology, OVC, University of Guelph, Guelph, ON, Canada, 6Dept of Mathematics and Statistics, University of Guelph, Guelph, ON, Canada, 7Canadian Dairy Network, Guelph, ON, Canada.

Milk fat globule (MFG) size in bovine milk is a trait of interest for selection due to its influence on the composition and nutritional quality of the milk as well as its technological properties. However, large-scale phenotyping is currently impractical. For a given fat content, milk with smaller MFG will have more membrane material, which has been proposed nutraceutical. The objective of this study was to examine the variation in MFG size and the potential of its prediction using mid-infrared spectral data. A total of 1,689 milk samples from 343 Canadian cows representing 4 dairy breeds from 44 herds across the provinces of Ontario, Alberta, and Quebec were collected during routine milk testing. Samples were divided and a certain portion was sent to a Canadian DHI laboratory and their spectral data were recorded using FOSS MIR machines. The MIR data for each sample contained 1,060 data points in the infrared range from 900 to 5,000 cm−1. The additional portion was analyzed for mean MFG size using integrated light scattering and reported as both a volume moment mean and surface moment mean. Mean values (±SD) for volume and surface moment means were 4.16 ± 0.51 µm and 3.49 ± 0.34 µm, respectively. The average MFG size was positively correlated with the percent fat in the sample (volume moment mean, r = 0.30; surface moment mean, r = 0.32). MFG size records were combined with their spectral data, and outliers and non-informative regions of the spectrum were removed. Records were randomly assigned to either the training or validation sets. Partial least squares regression method was utilized to predict MFG size. For volume moment mean, an R2cv of 0.52 and R2v of 0.41 were found. Surface moment mean equations had an R2cv of 0.55 and R2v of 0.52. At this time MFG size cannot be accurately quantified through MIR prediction, but it may be possible to identify milk samples with either small or large MFG.

Key Words: milk fat globule, dairy cattle, mid-infrared spectroscopy