利用畜禽全基因组SNP芯片的多性状基因组选择方法研究

With the availability of whole-genome high-density SNP (single nucleotide polymorphic) chips for domestic animals, genomic selection has become an international research interest for the worldwide animal breeding society. So far, most researchers in the area of genomic selection conducted their investigations only within the framework of single-trait genomic prediction model. Consequently, the informative records from the genetically correlated traits were not properly utilized by the existing genomic prediction models. Besides, a comprehensive investigation in the multi-trait genomic prediction model and its advantages over the corresponding single-trait model were rarely reported. However, multi-trait records were routinly included in the traditional genetic evaluation models, such as multi-trait animal model BLUP (best linear unbiased prediction), and their usefulness in improving the accuracy of prediction was clearly proven via the successful breeding practices in the past decades. Hereat, in this project, we will develop the arithmetic and the program for multi-trait genomic selection model, based on the existing single-trait genomic selection models and the trait-specific relationship matrix method. Additionally, the power of the proposed new model will be detailedly validated via both Monte Carlo method and cross-validation with a real dataset derived from a swine reference population. The aims of this project are to build a high-efficient, robust and high-accurate multi-trait genomic prediction model and program, and to provide more valuable experiences of multi-trait genomic selection for the animal breeding society.

随着畜禽全基因组高密度SNP芯片的市场化，基因组选择(Genomic Selection, GS)已成当前全球动物育种研究和应用的热点。目前的GS计算方法研究都仅对单性状遗传评估模型进行探讨，而忽略了对性状间的遗传相关信息，即对多性状记录的研究和利用。与单性状模型相比，传统的多性状动物模型BLUP可有效提高育种值估计的准确性，且已在长期的育种实践中所证明。而在GS研究领域，针对多性状模型的深入研究才刚刚起步且鲜有报道。因此，本研究将在单性状GS方法模型的基础上，整合性状特异亲缘关系信息，完整构建多性状GS算法并将其程序化，同时采用计算机蒙特卡罗模拟和猪群体全基因组实际数据交叉验证方法系统地评估多性状GS模型的理论和实际效果。通过本研究提出一套估计准确性高、计算稳定性强、运行效率高的多性状GS计算方法及程序，进一步完善GS方法体系，并为畜禽多性状GS的应用提供更多可参考的依据。

全基因组选择(Genomic Selection, GS)已成当前全球动物、植物育种研究及应用的热点，而多性状基因组选择方法是对常规基因组选择方法的有益扩展和补充。在本项目基金的支持下，我们开展了一系列基因组选择方法的创新性研究，并对多性状基因组选择的应用效果进行模拟和实际数据验证。主要内容及结果如下：（1）完成多性状模拟方法构建及软件开发，开发了基于突变漂移平衡的群体基因组数据模拟程序MTGPOPSIM,该软件可生成多性状群体遗传及表型数据，运行高效，操作灵活；（2）基因组选择计算方法研究取得突破，创新性地提出BLUP|GA和BayesBπ法，使用德国荷斯坦奶牛、水稻以及其他多个公共数据集对该方法进行验证均表明，所提出的方法可以有效利用性状遗传结构信息，提升基因组选择的准确性；（3）系统地开展多性状基因组选择应用效果模拟评估，对多性状GBLUP进行系统模拟验证，表明低遗传力性状在多性状GS遗传评估中可以有更多准确性提升，而遗传相关则与多性状GS准确性提升有明显关系；（4）构建猪、鸡基因组选择参考群，系统评估基因组选择在实际畜禽群体中的应用效果。在鸡验证群体中，多性状遗传评估均可明显提高GS准确性，提升幅度最大可达127%；（5）开展基因组选择一步法模拟对比研究，为基因组选择的实施提供技术支持。该项目研究共发表研究论文9篇，其中SCI论文6篇，获得软件著作权2项，培养研究生3名。