The genetic background of trait variability has captured the interest of ecologists and animal breeders because the genes that control it could be involved in buffering various environmental effects. Phenotypic variability of a given trait can be assessed by studying the heterogeneity of the residual variance, and the quantitative trait loci (QTL) that are involved in the control of this variability are described as variance QTL (vQTL). This study focuses on litter size (total number born, TNB) and its variability in a Large White pig population. The variability of TNB was evaluated either using a simple method, i.e. analysis of the log-transformed variance of residuals (LnVar), or the more complex double hierarchical generalized linear model (DHGLM). We also performed a single-SNP (single nucleotide polymorphism) genome-wide association study (GWAS). To our knowledge, this is only the second study that reports vQTL for litter size in pigs and the first one that shows GWAS results when using two methods to evaluate variability of TNB: LnVar and DHGLM. Based on LnVar, three candidate vQTL regions were detected, on Sus scrofa chromosomes (SSC) 1, 7, and 18, which comprised 18 SNPs. Based on the DHGLM, three candidate vQTL regions were detected, i.e. two on SSC7 and one on SSC11, which comprised 32 SNPs. Only one candidate vQTL region overlapped between the two methods, on SSC7, which also contained the most significant SNP. Within this vQTL region, two candidate genes were identified, ADGRF1, which is involved in neurodevelopment of the brain, and ADGRF5, which is involved in the function of the respiratory system and in vascularization. The correlation between estimated breeding values based on the two methods was 0.86. Three-fold cross-validation indicated that DHGLM yielded EBV that were much more accurate and had better prediction of missing observations than LnVar. The results indicated that the LnVar and DHGLM methods resulted in genetically different traits. Based on their validation, we recommend the use of DHGLM over the simpler method of log-transformed variance of residuals. These conclusions can be useful for future studies on the evaluation of the variability of any trait in any species.

中文翻译：

---

基于大白猪产仔数的变异表型和基因组背景评估

性状变异的遗传背景引起了生态学家和动物育种者的兴趣，因为控制它的基因可能参与缓冲各种环境影响。可以通过研究剩余方差的异质性来评估给定性状的表型变异性，并且将参与控制该变异性的数量性状基因座（QTL）描述为方差QTL（vQTL）。这项研究的重点是产仔数（出生总数，TNB）及其在大白猪群中的变异性。TNB 的可变性是使用一种简单的方法来评估的，即分析残差的对数变换方差 (LnVar)，或者使用更复杂的双层次广义线性模型 (DHGLM)。我们还进行了单 SNP（单核苷酸多态性）全基因组关联研究（GWAS）。据我们所知，这只是第二项报告猪窝产仔数 vQTL 的研究，也是第一项显示使用两种方法评估 TNB 变异性时 GWAS 结果的研究：LnVar 和 DHGLM。基于 LnVar，在 Sus scrofa 染色体 (SSC) 1、7 和 18 上检测到三个候选 vQTL 区域，其中包含 18 个 SNP。基于DHGLM，检测到3个候选vQTL区域，即SSC7上2个，SSC11上1个，包含32个SNP。在 SSC7 上，两种方法之间只有一个候选 vQTL 区域重叠，其中也包含最重要的 SNP。在这个 vQTL 区域内，确定了两个候选基因，参与大脑神经发育的 ADGRF1 和 ADGRF5，它与呼吸系统的功能和血管形成有关。基于两种方法估计的育种值之间的相关性为0.86。三重交叉验证表明，DHGLM 产生的 EBV 比 LnVar 更准确，并且对缺失观察的预测更好。结果表明，LnVar 和 DHGLM 方法产生了不同的遗传性状。基于他们的验证，我们建议使用 DHGLM，而不是更简单的残差对数变换方差方法。这些结论对于未来评估任何物种任何性状的变异性的研究很有用。三重交叉验证表明，DHGLM 产生的 EBV 比 LnVar 更准确，并且对缺失观察的预测更好。结果表明，LnVar 和 DHGLM 方法产生了不同的遗传性状。基于他们的验证，我们建议使用 DHGLM，而不是更简单的残差对数变换方差方法。这些结论对于未来评估任何物种任何性状的变异性的研究很有用。三重交叉验证表明，DHGLM 产生的 EBV 比 LnVar 更准确，并且对缺失观察的预测更好。结果表明，LnVar 和 DHGLM 方法产生了不同的遗传性状。基于他们的验证，我们建议使用 DHGLM，而不是更简单的残差对数变换方差方法。这些结论对于未来评估任何物种任何性状的变异性的研究很有用。