基于分子标记和系谱信息的虹鳟生长性状育种值估计方法研究

When the genetic information of markers with high linkage level, or the number of markers much larger than the number of individuals with phenotypic records, being considered in the best linear unbiased prediction (BLUP) of breeding value for growth trait based on pedigree in rainbow trout (Oncorhynchus mykiss), there are co-linear or excessive fixed effect factors in the mixed model equations. This kind of mixed model equations can not be solved by using the restricted maximum likelihood (REML) method. In this study, the least absolute shrinkage and selection operator (LASSO) and mixed model technologies would be employed to estimate breeding values with both molecular marker and pedigree information. The advantages of the proposed method on accuracy of genetic parameter estimation and computational efficiency would be demonstrated by using simulated data, and then the feasibility of this method would be validated in a real breeding population with clear pedigree for rainbow trout. The SLAF-seq (specific locus amplified fragment sequencing) technology would be used to develop high-throughput genetic markers, including large segment (eg. SSR) and SNP markers. The marker information would be incorporated into genetic evaluation for productive performances in rainbow trout, which could greatly improve the estimation accuracy of breeding values. Parental fish would be selected according to accurately estimated breeding values, which could enhance genetic progress of productive performances. Even if number of genetic markers available is enough for genomic selection, the estimation methods proposed here are also suitable to accurately predict genomic breeding value, because minor polygenes are considered in estimation of breeding values, except for major genes. The project could also be considered as a discussion about the methodology of breeding strategy for rainbow trout, especially on how to use all the effective genetic information in the breeding practice to improve the accuracy of selection. This study will also provide a reference to selective breeding for other fishes.

在将分子标记信息纳入基于系谱的虹鳟生长性状BLUP育种值估计时，当有较多标记间存在高度连锁或者标记数目超过带有表型记录的虹鳟个体数目时，模型中会存在共线性的或者过多的标记固定效应因子，这导致了约束极大似然法对此类过饱和混合线性模型方程组无法求解。本项目拟采用基于LASSO的求解过饱和混合模型方程组技术，建立结合分子标记和系谱信息的BLUP育种值估计方法；利用模拟数据验证该方法在遗传参数估计准确度和计算效率上的优势。以虹鳟选育群体生长性状为实际案例，通过SLAF-seq技术高通量获取试验群体个体分子标记基因型数据和通过生产性能定期测定获得的表型数据，估计生长性状个体育种值并解析分子标记和系谱信息分别对遗传方差的贡献率，同时验证本方法的有效性和可行性。本项目也是对虹鳟育种策略在方法学上的探讨，即如何将全部有效遗传信息应用在育种实践中，用以提高选种的准确性，本研究也将为其他鱼类遗传选育提供参考。

在将分子标记信息纳入基于系谱的虹鳟生长性状BLUP育种值估计时，当有较多标记间存在高度连锁或者标记数目超过带有表型记录的虹鳟个体数目时，模型中会存在共线性的或者过多的标记固定效应因子，这导致了约束极大似然法对此类过饱和混合线性模型方程组无法求解。针对上述问题，本研究设计了通过LASSO技术交叉验证压缩估计遗传效应非零的分子标记数目，联合LASSO和BLUP的结合分子标记信息和系谱信息的混合模型方程组方法，并命名为LASSO-LMM。编写了基于R语言和DMU软件包的模拟计算程序，然后分别采用基于系谱的传统BLUP方法、完全采用分子标记信息的方法以及结合分子标记和系谱信息的方法估计模拟群体个体育种值，比较分析了三种方法个体育种值的估计准确度，确认了LASSO-LMM方法的有效性。利用系谱结构清晰的虹鳟选育群体第四世代群体中，定期进行生长性能测定；并在上述群体中利用SLAF-seq技术获得35个全同胞家系组成的385尾个体的全基因组范围的SNPs基因型信息；采用转录组学方法建立了多基因表达与虹鳟屠体脂肪含量差异的相关性；利用获得的SNPs数据集估计了虹鳟屠体脂肪含量遗传参数和育种值。每个样品平均开发127, 546个SLAF标签，样本SLAF标签的平均测序深度为17.78×，达到项目申请书中预期的10×以上，共得到1, 144.53 M reads，平均GC含量为 43.97 %。通过生物信息学分析，获得155, 864个SLAF标签，其中多态性的SLAF标签共有123, 658个，共得到3, 199, 332个群体SNPs。继而，利用上述数据集估计了虹鳟鱼体脂肪含量个体育种值并剖分了各组分方差的贡献率。结果发现，该群体屠体脂肪含量基因组加性遗传力为 0.623 ± 0.175，存在一个小而显著的显性离差（1.428×10-7 ± 0.332）；基因组育种值与表型性状高度相关，皮尔森相关系数达到0.921。本研究结果有助于对虹鳟育种值估计方法的改进，也可以为其他鱼类种质资源评价和遗传改进提供技术参考。