Mitochondrial genomes differ from the nuclear genome and in humans it is known that mitochondrial variants contribute to genetic disorders. Prior to genomics, some livestock studies assessed the role of the mitochondrial genome but these were limited and inconclusive. Modern genome sequencing provides an opportunity to re-evaluate the potential impact of mitochondrial variation on livestock traits. This study first evaluated the empirical accuracy of mitochondrial sequence imputation and then used real and imputed mitochondrial sequence genotypes to study the role of mitochondrial variants on milk production traits of dairy cattle. The empirical accuracy of imputation from Single Nucleotide Polymorphism (SNP) panels to mitochondrial sequence genotypes was assessed in 516 test animals of Holstein, Jersey and Red breeds using Beagle software and a sequence reference of 1883 animals. The overall accuracy estimated as the Pearson’s correlation squared (R2) between all imputed and real genotypes across all animals was 0.454. The low accuracy was attributed partly to the majority of variants having low minor allele frequency (MAF < 0.005) but also due to variants in the hypervariable D-loop region showing poor imputation accuracy. Beagle software provides an internal estimate of imputation accuracy (DR2), and 10 percent of the total 1927 imputed positions showed DR2 greater than 0.9 (N = 201). There were 151 sites with empirical R2 > 0.9 (of 954 variants segregating in the test animals) and 138 of these overlapped the sites with DR2 > 0.9. This suggests that the DR2 statistic is a reasonable proxy to select sites that are imputed with higher accuracy for downstream analyses. Accordingly, in the second part of the study mitochondrial sequence variants were imputed from real mitochondrial SNP panel genotypes of 9515 Australian Holstein, Jersey and Red dairy cattle. Then, using only sites with DR2 > 0.900 and real genotypes, we undertook a genome-wide association study (GWAS) for milk, fat and protein yields. The GWAS mitochondrial SNP effects were not significant. The accuracy of imputation of mitochondrial genotypes from the SNP panel to sequence was generally low. The Beagle DR2 statistic enabled selection of sites imputed with higher empirical accuracy. We recommend building larger reference populations with mitochondrial sequence to improve the accuracy of imputing less common variants and ensuring that SNP panels include common variants in the D-loop region.

中文翻译：

---

线粒体序列变异：测试插补准确性及其与奶牛牛奶性状的关联


线粒体基因组与核基因组不同，在人类中已知线粒体变异会导致遗传性疾病。在基因组学出现之前，一些家畜研究评估了线粒体基因组的作用，但这些研究是有限的且不确定的。现代基因组测序提供了重新评估线粒体变异对牲畜性状的潜在影响的机会。本研究首先评估了线粒体序列推算的经验准确性，然后使用真实和推算的线粒体序列基因型来研究线粒体变异对奶牛产奶性状的作用。使用 Beagle 软件和 1883 只动物的序列参考，在 516 只荷斯坦、泽西和红色品种的测试动物中评估了单核苷酸多态性 (SNP) 组对线粒体序列基因型的估算准确性。所有动物的所有估算基因型和真实基因型之间的皮尔逊相关平方 (R2) 估计的总体准确度为 0.454。低准确度部分归因于大多数变异具有较低的次要等位基因频率 (MAF < 0.005)，但也归因于高变 D 环区域中的变异表现出较差的插补准确度。 Beagle 软件提供插补精度 (DR2) 的内部估计，1927 个插补位置总数中的 10% 显示 DR2 大于 0.9 (N = 201)。有 151 个位点具有经验 R2 > 0.9（测试动物中分离的 954 个变异），其中 138 个位点与 DR2 > 0.9 的位点重叠。这表明 DR2 统计数据是选择以更高准确度进行下游分析的位点的合理代理。 因此，在研究的第二部分中，线粒体序列变异是根据 9515 头澳大利亚荷斯坦奶牛、泽西奶牛和红色奶牛的真实线粒体 SNP 组基因型估算的。然后，仅使用 DR2 > 0.900 的位点和真实基因型，我们对牛奶、脂肪和蛋白质产量进行了全基因组关联研究 (GWAS)。 GWAS 线粒体 SNP 效应不显着。从 SNP 组合将线粒体基因型推算到序列的准确性普遍较低。 Beagle DR2 统计数据使得能够选择具有更高经验准确性的位点。我们建议使用线粒体序列构建更大的参考群体，以提高估算不太常见变异的准确性，并确保 SNP 组合包含 D 环区域中的常见变异。