Division of Animal Sciences, University of Missouri, Columbia, MO 65211-5300, USA.
BMC Genet. 2010 Apr 19;11:24. doi: 10.1186/1471-2156-11-24.
Molecular estimates of breeding value are expected to increase selection response due to improvements in the accuracy of selection and a reduction in generation interval, particularly for traits that are difficult or expensive to record or are measured late in life. Several statistical methods for incorporating molecular data into breeding value estimation have been proposed, however, most studies have utilized simulated data in which the generated linkage disequilibrium may not represent the targeted livestock population. A genomic relationship matrix was developed for 698 Angus steers and 1,707 Angus sires using 41,028 single nucleotide polymorphisms and breeding values were estimated using feed efficiency phenotypes (average daily feed intake, residual feed intake, and average daily gain) recorded on the steers. The number of SNPs needed to accurately estimate a genomic relationship matrix was evaluated in this population.
Results were compared to estimates produced from pedigree-based mixed model analysis of 862 Angus steers with 34,864 identified paternal relatives but no female ancestors. Estimates of additive genetic variance and breeding value accuracies were similar for AFI and RFI using the numerator and genomic relationship matrices despite fewer animals in the genomic analysis. Bootstrap analyses indicated that 2,500-10,000 markers are required for robust estimation of genomic relationship matrices in cattle.
This research shows that breeding values and their accuracies may be estimated for commercially important sires for traits recorded in experimental populations without the need for pedigree data to establish identity by descent between members of the commercial and experimental populations when at least 2,500 SNPs are available for the generation of a genomic relationship matrix.
由于选择准确性的提高和世代间隔的缩短，分子估计的育种值有望增加选择反应，特别是对于那些难以或昂贵记录或在生命后期测量的性状。已经提出了几种将分子数据纳入育种值估计的统计方法，然而，大多数研究都利用了模拟数据，其中产生的连锁不平衡可能无法代表目标家畜群体。利用 41,028 个单核苷酸多态性为 698 头安格斯公牛和 1,707 头安格斯公牛系谱建立了基因组关系矩阵，并利用在公牛身上记录的饲料效率表型（平均日采食量、剩余饲料摄入量和平均日增重）估计了育种值。在该群体中评估了准确估计基因组关系矩阵所需的 SNP 数量。
将结果与基于系谱的混合模型分析产生的估计值进行了比较，该分析基于 862 头安格斯公牛，其中有 34,864 个确定的父系亲属，但没有母系祖先。尽管在基因组分析中动物数量较少，但使用分子数和基因组关系矩阵估计 AFI 和 RFI 的加性遗传方差和育种值准确性相似。Bootstrap 分析表明，在牛中，需要 2,500-10,000 个标记才能稳健估计基因组关系矩阵。
本研究表明，当至少有 2,500 个 SNP 可用于生成基因组关系矩阵时，无需系谱数据即可在商业和实验群体成员之间建立血统关系，就可以为在实验群体中记录的性状估计商业上重要的公牛的育种值及其准确性。
