Animal Breeding and Genomics, Wageningen University & Research, Droevendaalsesteeg 1, P.O. Box 338, 6700 AH Wageningen, The Netherlands.
G3 (Bethesda). 2024 Nov 6;14(11). doi: 10.1093/g3journal/jkae205.
The ability to predict the outcome of selection and mating decisions enables breeders to make strategically better selection decisions. To improve genetic progress, those individuals need to be selected whose offspring can be expected to show high genetic variance next to high breeding values. Previously published approaches enable to predict the variance of descendants of 2 future generations for up to 4 founding haplotypes, or 2 outbred individuals, based on phased genotypes, allele effects, and recombination frequencies. The purpose of this study was to develop a general approach for the analytical calculation of the genetic variance in any future generation. The core development is an equation for the prediction of the variance of double haploid lines, under the assumption of no selection and negligible drift, stemming from an arbitrary number of founder haplotypes. This double haploid variance can be decomposed into gametic Mendelian sampling variances (MSVs) of ancestors of the double haploid lines allowing usage for non-double haploid genotypes that enables application in animal breeding programs as well as in plant breeding programs. Together with the breeding values of the founders, the gametic MSV may be used in new selection criteria. We present our idea of such a criterion that describes the genetic level of selected individuals in 4 generations. Since breeding programs do select, the assumption made for predicting variances is clearly violated, which decreases the accuracy of predicted gametic MSV caused by changes in allele frequency and linkage disequilibrium. Despite violating the assumption, we found high predictive correlations of our criterion to the true genetic level that was obtained by means of simulation for the "corn" and "cattle" genome models tested in this study (0.90 and 0.97). In practice, the genotype phases, genetic map, and allele effects all need to be estimated meaning inaccuracies in their estimation will lead to inaccurate variance prediction. Investigation of variance prediction accuracy when input parameters are estimated was not part of this study.
预测选择和交配决策结果的能力使饲养员能够做出更具策略性的选择决策。为了提高遗传进展，需要选择那些其后代可以预期具有高遗传方差和高育种值的个体。以前发表的方法可以根据分阶段基因型、等位基因效应和重组频率，预测多达 4 个起始单倍型或 2 个杂交个体的未来 2 代后代的方差。本研究的目的是开发一种通用方法，用于分析计算任何未来世代的遗传方差。核心开发是一个假设没有选择和可忽略的漂变的双单倍体系方差的预测方程，源于任意数量的起始单倍型。这个双单倍体方差可以分解为双单倍体系祖先的配子 Mendelian 抽样方差 (MSV)，允许用于非双单倍体基因型，这使得它能够在动物育种计划以及植物育种计划中应用。与创始人的育种值一起，配子 MSV 可用于新的选择标准。我们提出了这样一个标准的想法，该标准描述了 4 代中选定个体的遗传水平。由于育种计划确实在进行选择，因此用于预测方差的假设显然被违反，这会由于等位基因频率和连锁不平衡的变化而降低预测配子 MSV 的准确性。尽管违反了假设，但我们发现我们的标准与通过模拟获得的真实遗传水平之间存在很高的预测相关性，对于本研究中测试的“玉米”和“牛”基因组模型，该相关性分别为 0.90 和 0.97。在实践中，基因型相位、遗传图谱和等位基因效应都需要进行估计，这意味着它们的估计不准确将导致方差预测不准确。本研究未涉及输入参数估计时方差预测准确性的调查。
