Fleming A, Schenkel F S, Chen J, Malchiodi F, Ali R A, Mallard B, Sargolzaei M, Corredig M, Miglior F
Centre for Genetic Improvement of Livestock, Department of Animal Biosciences, University of Guelph, Guelph, ON, N1G 2W1, Canada.
Centre for Genetic Improvement of Livestock, Department of Animal Biosciences, University of Guelph, Guelph, ON, N1G 2W1, Canada.
J Dairy Sci. 2017 Mar;100(3):1640-1649. doi: 10.3168/jds.2016-11427. Epub 2017 Jan 11.
The objectives of this study were to investigate the sources of variation in milk fat globule (MFG) size in bovine milk and its prediction using mid-infrared (MIR) spectroscopy. Mean MFG size was measured in 2,076 milk samples from 399 Ayrshire, Brown Swiss, Holstein, and Jersey cows, and expressed as volume moment mean (D[4,3]) and surface moment mean (D[3,2]). The mid-infrared spectra of the samples and milk performance data were also recorded during routine milk recording and testing. The effects of breed, herd nested within breed, days in milk, season, milking period, age at calving, parity, and individual animal on the variation observed in MFG size were investigated. Breed, herd nested within breed, days in milk, season, and milking period significantly affected mean MFG size. Milk fat globule size was the largest at the beginning of lactation and subsequently decreased. Milk samples with the smallest MFG on average came from Holstein cows, and those with the largest were from Jersey and Brown Swiss cows. Partial least squares regression was used to predict MFG size from MIR spectra of samples with a calibration data set containing 2,034 and 2,032 samples for D[4,3] and D[3,2], respectively. Coefficients of determination of cross validation for D[4,3] and D[3,2] prediction models were 0.51 and 0.54, respectively. The associated ratio of performance deviation values were 1.43 and 1.48 for D[4,3] and D[3,2], respectively. With these models, individual mean MFG size could not be accurately predicted, but results may be sufficient to screen samples for having either small or large MFG on average. Significant but low correlations of D[4,3] and D[3,2] with milk fat yield were estimated (0.16 and 0.21, respectively). Significant and moderate Pearson correlation coefficients for fat percent with D[4,3] and D[3,2] were assessed (0.34 and 0.36, respectively). This correlation was greater between milk fat percentage and predicted MFG size than with measured MFG size with coefficients of 0.47 and 0.49 for D[4,3] and D[3,2], respectively. The MIR prediction equations are potentially overusing the correlation between fat and MFG size and exploiting the strong relationship between the MIR spectra and total milk fat. However, the predictions of MFG size are able to determine variation in mean globule size beyond what would be achieved just by looking at the correlation with fat production.
本研究的目的是调查牛乳中乳脂肪球（MFG）大小的变异来源，并利用中红外（MIR）光谱对其进行预测。对来自399头艾尔夏牛、瑞士褐牛、荷斯坦牛和泽西牛的2076份乳样的平均MFG大小进行了测量，并表示为体积矩平均直径（D[4,3]）和表面积矩平均直径（D[3,2]）。在常规的牛奶记录和检测过程中，还记录了这些样品的中红外光谱和牛奶性能数据。研究了品种、品种内的牛群、泌乳天数、季节、挤奶期、产犊时年龄、胎次和个体动物对MFG大小变异的影响。品种、品种内的牛群、泌乳天数、季节和挤奶期对平均MFG大小有显著影响。乳脂肪球大小在泌乳初期最大，随后减小。平均MFG最小的乳样来自荷斯坦牛，而最大的来自泽西牛和瑞士褐牛。使用偏最小二乘回归，根据校准数据集（分别包含2034个和2032个用于D[4,3]和D[3,2]的样品）的样品MIR光谱预测MFG大小。D[4,3]和D[3,2]预测模型的交叉验证决定系数分别为0.51和0.54。D[4,3]和D[3,2]的性能偏差值相关比率分别为1.43和1.48。利用这些模型，无法准确预测个体平均MFG大小，但结果可能足以筛选出平均MFG大小较小或较大的样品。估计D[4,3]和D[3,2]与乳脂肪产量的相关性显著但较低（分别为0.16和0.21）。评估了乳脂肪率与D[4,3]和D[3,2]的显著且中等的皮尔逊相关系数（分别为0.34和0.36）。乳脂肪百分比与预测的MFG大小之间的这种相关性大于与测量的MFG大小之间的相关性，D[4,3]和D[3,2]的相关系数分别为0.47和0.49。MIR预测方程可能过度利用了脂肪与MFG大小之间的相关性，并利用了MIR光谱与总乳脂肪之间的强关系。然而，MFG大小的预测能够确定平均球径的变异，这超出了仅通过观察与脂肪产量的相关性所能实现的范围。
