Single-step genomic evaluation in forest genetics

针对针叶树等具有复杂大基因组的非模式生物的基因组标记(SNPs)的出现，使基因组学在森林数量遗传学中的应用成为可能。传统的育种群体评估依赖于基于预期系谱的关系(A矩阵)来估计遗传参数。利用SNPs，个体间基于基因组的关系(G-矩阵)可以被估计出来，并作为个体-树混合模型中的A-矩阵的替代，通过最佳线性无偏预测(BLUP)来预测育种值。最近的植物和动物遗传评估表明，G-矩阵能够解释孟德尔分离和隐藏的谱系以及A-矩阵无法估计的历史谱系，因此具有优越性。在混合模型方程中拟合G-矩阵被称为基因组BLUP(GBLUP)，它已被证明在树木育种中是有效的；然而，树木育种群体非常大，有来自100个亲本的1000个后代，种植在多个地点。因此，GBLUP的实施受到基因分型成本(许多人)和后勤问题(跨地点采样)的阻碍。最近，人们提出了一种将A-矩阵和G-矩阵一步结合的新方法，该方法将传统的A-矩阵与G-矩阵同时使用，从而形成混合关系H-矩阵，该混合关系H-矩阵结合了多个非分型个体的A-矩阵和分型个体的子集的G-矩阵。H-矩阵可以被视为利用系谱关系从遗传型个体到非遗传型个体的遗传优点的投影。因此，当A矩阵作为跨越个体和父母的桥梁时，在混合方法中产生额外的信息，从而最终促进在BLUP分析期间更好地利用信息，从而产生更可靠和准确的遗传参数。H矩阵在动物育种中很常见，因为基因×环境互作(GxE)很小，这种情况不适用于树木测试，因为GxE很常见。由于H-矩阵的实现需要基因分型(子集)和非基因分型(整个子代测试)，因此必须回答几个问题，包括：1)构建G-矩阵需要多少个SNP？2)基因分型和非基因分型之间的平衡是什么？3)基因分型可以限制在特定的位置吗？以及4)GxE对估计的遗传参数有什么影响？在这里，我计划在不列颠哥伦比亚省的白云杉40年树龄测试计划上实施H矩阵，该计划非常适合提供可靠的生长和木材属性。**

The availability of genomic marker (SNPs) for non-model organisms with complex and large genome such as conifers, made it possible to infuse genomics in forest quantitative genetics. Traditional evaluation of breeding populations relied on the expected pedigree-based relationship (A-matrix) for estimating the genetic parameters. With SNPs, the genomic-based relationship (G-matrix) between individuals can be estimated and used as a substitute to the A-matrix in the individual-tree mixed model to predict breeding values by the Best Linear Unbiased Prediction (BLUP). Recent plant and animal genetic evaluations have demonstrated the superiority of G-matrix for its ability to account for the Mendelian segregation and hidden pedigree as well as historical pedigree that cannot be estimated with the A-matrix. Fitting the G-matrix in the mixed model equations is known as the genomic BLUP (GBLUP) which has proven to be effective in tree breeding; however, the tree breeding populations are exceedingly large, with 1000s of progenies from 100s parents, planted over multiple sites. Thus, GBLUP implementation is hampered by genotyping costs (many individuals) and logistical issues (sampling across sites). Recently, a novel approach combines both the A- and G-matrices in a single-step was proposed, where the traditional A-matrix is concurrently utilized with the G-matrix, thus forming a blended relationship H-matrix combines the A-matrix of many non-genotyped individuals and G-matrix of a subset of genotyped individuals. The H-matrix can be seen as a projection of genetic merit from genotyped to non-genotyped individuals using pedigree relationships. Thus, additional information is generated in the blended approach as the A-matrix acts like bridges across individuals and parents, thus ultimately facilitating better information utilization during the BLUP analysis, resulting in the generation of more reliable and accurate genetic parameters. The H-matrix is common in animal breeding as genotype x environment interaction (GxE) is minimal, a situation does not apply to tree testing as GxE is common. Since the implementation of the H-matrix requires genotyped (subset) and non-genotyped (the entire progeny test), then several questions must be answered, including; 1) what is the number of SNPs needed for constructing the G-matrix?, 2) what is the balance between genotyped and non-genotyped?, 3) can genotyping be limited to specific sites?, and 4) what is the impact of GxE on the estimated genetic parameters? Here, I plan to implement the H-matrix on the British Columbia's white spruce 40 year-old testing program which is ideal for providing reliable growth and wood attributes.**