基于测定日模型建立中国荷斯坦奶牛基因组选择新方法

Genomic selection has been acting as the state-of-the-arts technology for molecular assisted selection in animal breeding. Especially in dairy cattle, genomic evaluation is currently becoming the leading way for genetic evaluation and individual selection, which is expected to be the substitution of traditional progeny test approach in a few years. So far, a typical genomic evaluation in dairy cattle involves several steps. This "multiple step" strategy severely depends on many parameters, assumptions, and particularly pseudo-observations, such as daughter deviations or deregressed proofs. The features of "multiple step" strategy entail several limitations in statistics, lowering the accuracy and unbiasness of estimated genomic breeding value (gEBV). To address this concern on genomic selection, an improved "single step" method has been proposed to perform genomic evaluation. Through "single step" strategy, genomic selection and routine genetic evaluation can be integrated into an unified framework, and the traditional EBV and gEBV can be obtained simultaneously under such situation. In present project, a novel genomic breeding value estimation approach is developed focusing on test-day milk production traits in dairy cattle. In our method, a random regression test-day model is constructed considering both pedigree and genomic relationship among individuals. Under this model, gEBV can be estimated through random regression coefficient of additive effect of each individual for test-day milk production traits. The performance of the proposed method will be further validated via extensive simulation as well as empirical data analyses in Chinese Holstein. Our study goal is to enhance the "single step" genomic evaluation strategy for process longitudinal phenotypic observations. The success of implementation of this project will eventually develop an optimal "single step " genomic selection program for test-day milk production traits in Chinese Holstein. The proposed novel strategy will not only promote the progress of theoretical investigation in genomic selection, but also lay a solid technical foundation for performing genomic selection project in Chinese Holstein in practice.

基因组选择是当前畜禽分子选育的前沿技术。对于奶牛育种，基因组选择正逐步代替常规方法成为遗传评估的重要手段。在奶牛基因组选择中，利用"多步"法策略进行gEBV估计是当前国际主流。"多步"法依赖于性状"伪表型"信息，存在明显的理论缺陷。最近提出的"一步"法进行gEBV估计的改进策略，可将基因组选择和常规遗传评估纳入统一框架实现一步估计，克服了现有方法的不足。本研究首次提出将随机回归测定日模型引入 "一步"法的新策略，旨在克服现有 "一步"法不能利用奶牛测定日表型资料的不足，通过估计基因组育种值随机回归参数的原理进行测定日产奶性状gEBV一步估计。研究采用系统模拟和中国荷斯坦奶牛实际群体分析的手段，对新方法估计效果进行分析和验证。本项目的实施，将建立测定日性状基因组选择"一步"分析的优化方法，从而进一步完善基因组选择"一步"法的理论体系，为我国奶牛基因组选择的优化实施提供技术保障。

产奶性状是奶牛最重要的经济性状，提高奶牛产奶性状选择的准确性、加大遗传进展具有重要的经济与社会意义。在奶牛常规遗传评估中，常用的模型是随机回归测定日模型，在逐渐成为主流遗传评估方法的基因组选择中，“一步”法比“两步”法更具有优势。本项目提出的基于随机回归测定日模型的基因组选择“一步”法，旨在结合二者的优势，从而提高育种值估计的准确性与无偏性。研究内容主要包括：1）建立基于随机回归测定日模型的基因组选择“一步”法的理论技术体系；2）构建可以实现包括这一方法在内的、基于不同模型的大型遗传评估的通用平台；3）通过模拟不同参数组合的数据，对所提出的方法进行验证；4）在实际数据中对这一方法进行验证。目前，本项目的研究内容以全部圆满完成，并在相关领域进行了拓展研究，取得了如下成果：1）建立了测定日性状基因组选择“一步”分析的优化方法，完善了基因组选择理论技术体系；2）构建了可快速、灵活进行大规模遗传评估的计算平台，为遗传评估的实施提供了基础；3）通过模拟研究，验证了所提出的方法在育种值估计的准确性和无偏性上均具有优势；4）自本项目立项起，连续多年参与中国荷斯坦牛群体的基因组遗传评估，显著提高了中国荷斯坦奶牛的遗传改良速度；5）基因组选择方面，在实现贝叶斯经典方法的基础上，对其进行了改进，提出了BayesA+、BayesB+、BayesCπ+和多性状贝叶斯前相关模型，模拟与实际数据验证表明这些方法可以一定程度上提高育种值估计的准确性；6）测定日模型方面，将随机回归测定日模型引入全基因组关联分析，提出了测定日数据全基因组关联分析的新方法，模拟研究表明，这一方法在控制假阳性率的同时，可以提高检验功效；7）提出了通过将个体遗传效应反推到SNP上位效应以实现快速检验的全基因组上位分析方法，模拟研究表明，这一方法可以更有效的控制假阳性率，并大幅提高了计算速度。综上，本项目的研究成果在育种理论发展与实践中均具有重要意义。