Department of Agricultural Sciences, University of Sassari, 07100 Sassari, Italy.
Department of Agricultural Sciences, University of Sassari, 07100 Sassari, Italy; Department of Animal and Dairy Science, University of Georgia, Athens 30602.
J Dairy Sci. 2021 Apr;104(4):5079-5094. doi: 10.3168/jds.2020-19087. Epub 2021 Jan 28.
Fatty acid (FA) profile is one of the most important aspects of the nutritional properties of milk. The FA content in milk is affected by several factors such as diet, physiology, environment, and genetics. Recently, principal component analysis (PCA) and multivariate factor analysis (MFA) have been used to summarize the complex correlation pattern of the milk FA profile by extracting a reduced number of new variables. In this work, the milk FA profile of a sample of 993 Sarda breed ewes was analyzed with PCA and MFA to compare the ability of these 2 multivariate statistical techniques in investigating the possible existence of latent substructures, and in studying the influence of physiological and environmental effects on the new extracted variables. Individual scores of PCA and MFA were analyzed with a mixed model that included the fixed effects of parity, days in milking, lambing month, number of lambs born, altitude of flock location, and the random effect of flock nested within altitude. Both techniques detected the same number of latent variables (9) explaining 80% of the total variance. In general, PCA structures were difficult to interpret, with only 4 principal components being associated with a clear meaning. Principal component 1 in particular was the easiest to interpret and agreed with the interpretation of the first factor, with both being associated with the FA of mammary origin. On the other hand, MFA was able to identify a clear structure for all the extracted latent variables, confirming the ability of this technique to group FA according to their function or metabolic origin. Key pathways of the milk FA metabolism were identified as mammary gland de novo synthesis, ruminal biohydrogenation, desaturation performed by stearoyl-coenzyme A desaturase enzyme, and rumen microbial activity, confirming previous findings in sheep and in other species. In general, the new extracted variables were mainly affected by physiological factors as days in milk, parity, and lambing month; the number of lambs born had no effect on the new variables, and altitude influenced only one principal component and factor. Both techniques were able to summarize a larger amount of the original variance into a reduced number of variables. Moreover, factor analysis confirmed its ability to identify latent common factors clearly related to FA metabolic pathways.
脂肪酸（FA）组成是牛奶营养特性的最重要方面之一。牛奶中的 FA 含量受饮食、生理、环境和遗传等多种因素的影响。最近，主成分分析（PCA）和多变量因子分析（MFA）已被用于通过提取较少的新变量来总结牛奶 FA 组成的复杂相关模式。在这项工作中，分析了 993 只萨拉丁品种母羊的牛奶 FA 组成，采用 PCA 和 MFA 进行比较，以比较这 2 种多元统计技术在研究潜在亚结构的存在以及研究生理和环境效应对新提取变量的影响方面的能力。个体的 PCA 和 MFA 得分采用包含固定效应（胎次、泌乳天数、产羔月份、产羔数、羊群所在海拔高度）和随机效应（嵌套在海拔高度内的羊群）的混合模型进行分析。两种技术都检测到相同数量的潜在变量（9 个），解释了总方差的 80%。总体而言，PCA 结构难以解释，只有 4 个主成分与明确的含义相关。特别是主成分 1 最容易解释，与第一因子的解释一致，两者都与乳腺来源的 FA 相关。另一方面，MFA 能够为所有提取的潜在变量确定清晰的结构，证实了该技术根据其功能或代谢起源对 FA 进行分组的能力。牛奶 FA 代谢的关键途径被确定为乳腺从头合成、瘤胃生物氢化、硬脂酰辅酶 A 去饱和酶进行的去饱和作用以及瘤胃微生物活性，这与绵羊和其他物种的先前发现一致。总体而言，新提取的变量主要受生理因素（泌乳天数、胎次和产羔月份）的影响；产羔数对新变量没有影响，海拔仅影响一个主成分和因子。两种技术都能够将更多的原始方差总结为较少的变量。此外，因子分析证实了其能够清晰地识别与 FA 代谢途径明显相关的潜在共同因素的能力。
