The Roslin Institute and Royal (Dick) School of Veterinary Studies, University of Edinburgh, Midlothian, EH25 9RG, UK.
Centre for Tropical Livestock Genetics and Health (CTLGH), Roslin Institute, University of Edinburgh, Easter Bush Campus, Midlothian, EH25 9RG, UK.
Genet Sel Evol. 2022 Sep 4;54(1):58. doi: 10.1186/s12711-022-00751-5.
In cattle, genome-wide association studies (GWAS) have largely focused on European or Asian breeds, using genotyping arrays that were primarily designed for European cattle. Because there is growing interest in performing GWAS in African breeds, we have assessed the performance of 23 commercial bovine genotyping arrays for capturing the diversity across African breeds and performing imputation. We used 409 whole-genome sequences (WGS) spanning global cattle breeds, and a real cohort of 2481 individuals (including African breeds) that were genotyped with the Illumina high-density (HD) array and the GeneSeek bovine 50 k array.
We found that commercially available arrays were not effective in capturing variants that segregate among African indicine animals. Only 6% of these variants in high linkage disequilibrium (LD) (r > 0.8) were on the best performing arrays, which contrasts with the 17% and 25% in African and European taurine cattle, respectively. However, imputation from available HD arrays can successfully capture most variants (accuracies up to 0.93), mainly when using a global, not continent-specific, reference panel, which partially reflects the unusually high levels of admixture on the continent. When considering functional variants, the GGPF250 array performed best for tagging WGS variants and imputation. Finally, we show that imputation from low-density arrays can perform almost as well as HD arrays, if a two-stage imputation approach is adopted, i.e. first imputing to HD and then to WGS, which can potentially reduce the costs of GWAS.
Our results show that the choice of an array should be based on a balance between the objective of the study and the breed/population considered, with the HD and BOS1 arrays being the best choice for both taurine and indicine breeds when performing GWAS, and the GGPF250 being preferable for fine-mapping studies. Moreover, our results suggest that there is no advantage to using the indicus-specific arrays for indicus breeds, regardless of the objective. Finally, we show that using a reference panel that better represents global bovine diversity improves imputation accuracy, particularly for non-European taurine populations.
在牛中，全基因组关联研究（GWAS）主要集中在欧洲或亚洲品种上，使用的基因分型阵列主要是为欧洲牛设计的。由于人们对在非洲品种中进行 GWAS 的兴趣日益浓厚，我们评估了 23 种商业牛基因分型阵列在捕获非洲品种多样性和进行 imputation 方面的性能。我们使用了 409 个全基因组序列（WGS），涵盖了全球牛品种，以及一个由 2481 个个体组成的真实队列（包括非洲品种），这些个体使用 Illumina 高密度（HD）阵列和 GeneSeek 牛 50k 阵列进行了基因分型。
我们发现，商业上可用的阵列在捕获非洲指示动物中分离的变体方面效果不佳。在高连锁不平衡（LD）（r>0.8）中，只有 6%的这些变体位于表现最好的阵列上，这与非洲和欧洲的 tauine 牛中的 17%和 25%形成对比。然而，来自可用的 HD 阵列的 imputation 可以成功捕获大多数变体（准确性高达 0.93），主要是使用全球而非特定于大陆的参考面板时，这部分反映了非洲大陆异常高水平的混合。在考虑功能变体时，GGPF250 阵列在标记 WGS 变体和 imputation 方面表现最佳。最后，我们表明，如果采用两阶段 imputation 方法，即首先 imputation 到 HD，然后 imputation 到 WGS，则来自低密度阵列的 imputation 几乎可以与 HD 阵列一样好，这可能会降低 GWAS 的成本。
我们的结果表明，阵列的选择应基于研究目标和考虑的品种/群体之间的平衡，对于 tauine 和 indicine 品种，HD 和 BOS1 阵列是 GWAS 的最佳选择，而 GGPF250 则更适合精细映射研究。此外，我们的结果表明，无论目标如何，对于 indicine 品种，使用 indicus 特异性阵列没有优势。最后，我们表明，使用更好地代表全球牛多样性的参考面板可以提高 imputation 准确性，特别是对于非欧洲的 tauine 群体。
