Alemu Setegn Worku, Lopdell Thomas J, Trevarton Alexander J, Snell Russell G, Littlejohn Mathew D, Garrick Dorian J
AL Rae Centre for Genetics and Breeding, Massey University, 10 Bisley Drive, Hamilton, 3240, New Zealand.
Invermay Agricultural Centre, AgResearch Limited, Mosgiel, New Zealand.
Genet Sel Evol. 2025 Apr 11;57(1):20. doi: 10.1186/s12711-025-00966-2.
Genomic selection, typically employing genetic markers from SNP chips, is routine in modern dairy cattle breeding. This study assessed the impact of functional sequence variants on genomic prediction accuracy relative to 50 k SNP chip markers for fat percent, protein percent, milk volume, fat yield, and protein yield in lactating dairy cattle. The functional variants were identified through GWAS, RNA-seq, Histone modification ChIP-seq, ATAC-seq, or were coding variants. The genomic prediction accuracy obtained using each class of functional variants was compared with matched numbers of SNPs randomly selected from the Illumina 50 k SNP chip.
The investigation revealed that variants identified by GWAS or RNA-seq, significantly improved the prediction accuracy across all five traits. Contributions from ChIP-seq, ATAC-seq, and coding variants varied. Some variants identified using ChIP-seq showed marked improvements, while others reduced accuracy in protein yield predictions. Relative to a matched number of 32,595 SNPs from the SNP chip, pooling all the functional variants demonstrated prediction accuracy increases of 1.76% for fat percent, 2.97% for protein percent, 0.51% for milk volume, and 0.26% for fat yield, but with a slight decrease of 0.43% in protein yield.
The study demonstrates that functional variants can improve prediction accuracy relative to equivalent numbers of variants from a generic SNP panel, with percent traits showing more significant gains than yield traits. The main advantage of using functional variants for genomic prediction was achievement of comparable accuracy using a smaller, more selective set of loci. This is particularly evident in trait-specific scenarios. Our findings indicate that specific combinations of functional variants comprising 16 k variants can achieve genomic prediction accuracy comparable to employing a standard panel of twice the size (32.6 k), especially for percent traits. This highlights the potential for the development of more efficient, trait-focused SNP panels utilizing functional variants.
基因组选择通常采用SNP芯片中的遗传标记，在现代奶牛育种中已成为常规方法。本研究评估了功能序列变异相对于50k SNP芯片标记对泌乳奶牛脂肪百分比、蛋白质百分比、产奶量、脂肪产量和蛋白质产量的基因组预测准确性的影响。功能变异通过全基因组关联研究（GWAS）、RNA测序（RNA-seq）、组蛋白修饰染色质免疫沉淀测序（ChIP-seq）、转座酶可接近染色质测序（ATAC-seq）鉴定，或为编码变异。将使用每类功能变异获得的基因组预测准确性与从Illumina 50k SNP芯片中随机选择的匹配数量的单核苷酸多态性（SNP）进行比较。
研究表明，通过GWAS或RNA-seq鉴定的变异显著提高了所有五个性状的预测准确性。ChIP-seq、ATAC-seq和编码变异的贡献各不相同。一些通过ChIP-seq鉴定的变异显示出显著改善，而其他变异则降低了蛋白质产量预测的准确性。相对于来自SNP芯片的32595个匹配SNP数量，汇集所有功能变异显示，脂肪百分比的预测准确性提高了1.76%，蛋白质百分比提高了2.97%，产奶量提高了0.51%，脂肪产量提高了0.26%，但蛋白质产量略有下降0.43%。
该研究表明，相对于通用SNP面板中同等数量的变异，功能变异可以提高预测准确性，百分比性状的提高比产量性状更显著。使用功能变异进行基因组预测的主要优势是使用更小、更具选择性的位点集可实现相当的准确性。这在特定性状的情况下尤为明显。我们的研究结果表明，包含16k变异的功能变异的特定组合可以实现与使用两倍大小（32.6k）的标准面板相当的基因组预测准确性，尤其是对于百分比性状。这突出了利用功能变异开发更高效、针对特定性状的SNP面板的潜力。
