PDF(Pubmed)
Abstract：
BACKGROUND: Genomic prediction of breeding values (GP) has been adopted in evolutionary genomic studies to uncover microevolutionary processes of wild populations or improve captive breeding strategies. While recent evolutionary studies applied GP with individual single nucleotide polymorphism (SNP), haplotype-based GP could outperform individual SNP predictions through better capturing the linkage disequilibrium (LD) between the SNP and quantitative trait loci (QTL). This study aimed to evaluate the accuracy and bias of haplotype-based GP of immunoglobulin (Ig) A (IgA), IgE, and IgG against Teladorsagia circumcincta in lambs of an unmanaged sheep population (Soay breed) based on Genomic Best Linear Unbiased Prediction (GBLUP) and five Bayesian [BayesA, BayesB, BayesCπ, Bayesian Lasso (BayesL), and BayesR] methods.
RESULTS: The accuracy and bias of GPs using SNP, haplotypic pseudo-SNP from blocks with different LD thresholds (0.15, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, and 1.00), or the combinations of pseudo-SNPs and non-LD clustered SNPs were obtained. Across methods and marker sets, higher ranges of genomic estimated breeding values (GEBV) accuracies were observed for IgA (0.20 to 0.49), followed by IgE (0.08 to 0.20) and IgG (0.05 to 0.14). Considering the methods evaluated, up to 8% gains in GP accuracy of IgG were achieved using pseudo-SNPs compared to SNPs. Up to 3% gain in GP accuracy for IgA was also obtained using the combinations of the pseudo-SNPs with non-clustered SNPs in comparison to fitting individual SNP. No improvement in GP accuracy of IgE was observed using haplotypic pseudo-SNPs or their combination with non-clustered SNPs compared to individual SNP. Bayesian methods outperformed GBLUP for all traits. Most scenarios yielded lower accuracies for all traits with an increased LD threshold. GP models using haplotypic pseudo-SNPs predicted less-biased GEBVs mainly for IgG. For this trait, lower bias was observed with higher LD thresholds, whereas no distinct trend was observed for other traits with changes in LD.
CONCLUSIONS: Haplotype information improves GP performance of anti-helminthic antibody traits of IgA and IgG compared to fitting individual SNP. The observed gains in the predictive performances indicate that haplotype-based methods could benefit GP of some traits in wild animal populations.
RESULTS: The accuracy and bias of GPs using SNP, haplotypic pseudo-SNP from blocks with different LD thresholds (0.15, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, and 1.00), or the combinations of pseudo-SNPs and non-LD clustered SNPs were obtained. Across methods and marker sets, higher ranges of genomic estimated breeding values (GEBV) accuracies were observed for IgA (0.20 to 0.49), followed by IgE (0.08 to 0.20) and IgG (0.05 to 0.14). Considering the methods evaluated, up to 8% gains in GP accuracy of IgG were achieved using pseudo-SNPs compared to SNPs. Up to 3% gain in GP accuracy for IgA was also obtained using the combinations of the pseudo-SNPs with non-clustered SNPs in comparison to fitting individual SNP. No improvement in GP accuracy of IgE was observed using haplotypic pseudo-SNPs or their combination with non-clustered SNPs compared to individual SNP. Bayesian methods outperformed GBLUP for all traits. Most scenarios yielded lower accuracies for all traits with an increased LD threshold. GP models using haplotypic pseudo-SNPs predicted less-biased GEBVs mainly for IgG. For this trait, lower bias was observed with higher LD thresholds, whereas no distinct trend was observed for other traits with changes in LD.
CONCLUSIONS: Haplotype information improves GP performance of anti-helminthic antibody traits of IgA and IgG compared to fitting individual SNP. The observed gains in the predictive performances indicate that haplotype-based methods could benefit GP of some traits in wild animal populations.
摘要：
背景：进化基因组研究已采用育种值（GP）的基因组预测来揭示野生种群的微观进化过程或改善圈养育种策略。虽然最近的进化研究应用了具有个体单核苷酸多态性（SNP）的GP，通过更好地捕获SNP和数量性状基因座（QTL）之间的连锁不平衡（LD），基于单倍型的GP可以优于单个SNP预测。本研究旨在评估基于单倍型的免疫球蛋白（Ig）A（IgA）GP的准确性和偏倚，IgE,基于基因组最佳线性无偏预测（GBLUP）和五个贝叶斯[贝叶斯A，贝叶斯B,贝叶斯Cπ，贝叶斯套索(BayesL)，和BayesR]方法。
结果：使用SNP的GP的准确性和偏差，来自具有不同LD阈值（0.15、0.2、0.3、0.4、0.5、0.6、0.7、0.8、0.9和1.00）的区块的单倍型伪SNP，或获得伪SNP和非LD聚类SNP的组合。跨方法和标记集，观察到IgA的基因组估计育种值(GEBV)准确度范围更高(0.20至0.49)，其次是IgE（0.08至0.20）和IgG（0.05至0.14）。考虑到评估的方法，与SNP相比,使用假SNP,IgG的GP准确度提高高达8%。与拟合个体SNP相比,使用伪SNP与非聚类SNP的组合也获得了IgA的GP准确度的高达3%的增益。与个体SNP相比,使用单倍型假SNP或其与非成簇SNP的组合未观察到IgE的GP准确性的改善。贝叶斯方法在所有性状上都优于GBLUP。大多数情况下，LD阈值增加的所有性状的准确性较低。使用单倍型假SNP的GP模型预测主要针对IgG的较少偏倚的GEBV。对于这个特点，较低的偏差观察到较高的LD阈值，而其他性状随着LD的变化没有观察到明显的趋势。
结论：与拟合单个SNP相比，单倍型信息改善了IgA和IgG的抗蠕虫抗体性状的GP性能。观察到的预测性能增益表明，基于单倍型的方法可以使野生动物种群中某些性状的GP受益。
结果：使用SNP的GP的准确性和偏差，来自具有不同LD阈值（0.15、0.2、0.3、0.4、0.5、0.6、0.7、0.8、0.9和1.00）的区块的单倍型伪SNP，或获得伪SNP和非LD聚类SNP的组合。跨方法和标记集，观察到IgA的基因组估计育种值(GEBV)准确度范围更高(0.20至0.49)，其次是IgE（0.08至0.20）和IgG（0.05至0.14）。考虑到评估的方法，与SNP相比,使用假SNP,IgG的GP准确度提高高达8%。与拟合个体SNP相比,使用伪SNP与非聚类SNP的组合也获得了IgA的GP准确度的高达3%的增益。与个体SNP相比,使用单倍型假SNP或其与非成簇SNP的组合未观察到IgE的GP准确性的改善。贝叶斯方法在所有性状上都优于GBLUP。大多数情况下，LD阈值增加的所有性状的准确性较低。使用单倍型假SNP的GP模型预测主要针对IgG的较少偏倚的GEBV。对于这个特点，较低的偏差观察到较高的LD阈值，而其他性状随着LD的变化没有观察到明显的趋势。
结论：与拟合单个SNP相比，单倍型信息改善了IgA和IgG的抗蠕虫抗体性状的GP性能。观察到的预测性能增益表明，基于单倍型的方法可以使野生动物种群中某些性状的GP受益。
