中国荷斯坦牛繁殖性状的基因组预测效果比较  

Comparisons of Genomic Predictions for Fertility Traits in Chinese Holstein Cattle

作　　者：师睿[1] 苏国生[2] 陈紫薇 李想[1] 罗汉鹏 刘林[3] 郭刚 张毅[1] 王雅春[1] 张胜利[1] 张勤[1,5] SHI Rui;SU Guosheng;CHEN Ziwei;LI Xiang;LUO Hanpeng;LIU Lin;GUO Gang;ZHANG Yi;WANG Yachun;ZHANG Shengli;ZHANG Qin(College of Animal Science and Technology,China Agricultural University,Beijing 100193,China;Arhus University,Tjele DK-8830,Denmark;Beijing Dairy Cattle Center,Beijing 100192,China;Beijing Sunlon Livestock Development Co.Ltd.,Beijing 100176,China;College of Animal Science and Technology,Shandong Agricultural University,Tai’an 271018,China)

机构地区：[1]中国农业大学动物科技学院,北京100193 [2]丹麦奥胡斯大学,切勒DK-8830 [3]北京奶牛中心,北京100192 [4]北京首农畜牧发展有限公司,北京100176 [5]山东农业大学动物科技学院,泰安271018

出　　处：《畜牧兽医学报》2022年第9期2944-2954,共11页ACTA VETERINARIA ET ZOOTECHNICA SINICA

基　　金：财政部和农业农村部:国家现代农业产业技术体系(CARS-36);长江学者和创新团队发展计划(IRT_15R62);国家留学基金委(留金美[2019]13043);北京三元种业科技股份有限公司自立科研课题(SYZYZ20190005)。

摘　　要：旨在比较不同方法对中国荷斯坦牛繁殖性状的基因组预测效果,选择最佳的基因组预测方法及信息矩阵权重组合(τ和ω)用于实际育种。本研究利用北京地区33个牧场1998—2020年荷斯坦牛群繁殖记录,分析了3个重要繁殖性状:产犊至首次配种间隔(ICF)、青年牛配种次数(NSH)和成母牛配种次数(NSC)共98 483~197 764条表型数据。同时收集了8 718头母牛和3 477头公牛的基因芯片数据,根据具有芯片数据的牛群结构划分为公牛验证群和母牛验证群。随后,通过BLUPF90软件的AIREMLF90和BLUPF90模块利用最佳线性无偏预测(BLUP)、基因组最佳线性无偏预测(GBLUP)和一步法(ssGBLUP)对3个性状进行基因组预测,不同方法的预测效果根据准确性和无偏性来评估。结果表明,3个繁殖性状均为低遗传力性状(0.03~0.08);ssGBLUP方法中,各性状信息矩阵的权重取值能够在一定程度上提升基因组预测的效果;ICF、NSH和NSC在母牛验证群下的最佳权重取值分别为:τ=1.3和ω=0,τ=0.5和ω=0.4以及τ=0.5和ω=0;在公牛验证群下最优权重组合分别为:τ=1.5和ω=0,τ=1.3和ω=0.8以及τ=0.5和ω=0;基于最佳权重的ssGBLUP方法准确性较BLUP和GBLUP方法准确性分别提升了0.10~0.39和0.08~0.15,且无偏性最接近于1。综上,使用最佳权重组合的ssGBLUP时,各性状基因组预测结果具有较高准确性和无偏性,建议作为中国荷斯坦牛繁殖性状基因组选择方法。This study aimed to compare different methods for genomic predictions of fertility traits in Chinese Holstein cows, and to select the optimal method and combination of scaling factors(τ and ω) for practical breeding program. The raw fertility data were collected from 33 Holstein dairy farms in Beijing, covering the period from 1998 to 2020. The data included a total of 98 483 to 197 764 phenotypic records for the traits of interval from calving to first service(ICF), number of services for heifers(NSH) and number of services for cows(NSC). Meanwhile, genotypes of 8 718 cows and 3 477 bulls were collected. Both bull validation population and cow validation population were generated based on the structure of populations with genotypes data. Afterwards, best linear unbiased prediction(BLUP), genomic best linear unbiased prediction(GBLUP) and single-step genomic best linear unbiased prediction(ssGBLUP) were used for the predictions of the 3 target traits by using AIREMLF90 and BLUPF90 modules in BLUPF90 software. Prediction accuracy and unbiasedness were calculated to evaluate the effect of predictions. The results indicated that all 3 traits had low heritability(0.03-0.08), and weights of each trait information matrix in ssGBLUP methods had impacts on prediction accuracy and unbiasedness. The optimal combination of scaling factors for ICF, NSH and NSC were τ=1.3 and ω=0, τ=0.5 and ω=0.4, and τ=0.5 and ω=0, respectively based on cow validation population, while the combinations were τ=1.5 and ω=0, τ=1.3 and ω=0.8, and τ=0.5 and ω=0, respectively for bull validation population. The accuracy of the ssGBLUP method based on the best weight was 0.10-0.39 and 0.08-0.15 higher than that of the BLUP and GBLUP methods, respectively, and the unbiasedness was closest to 1. In conclusion, ssGBLUP with optimized weights produced the most accurate and unbiased predictions, thus it is suggested to be used in practical genomic selection program for fertility traits of Chinese Holstein cattle.

关 键 词：奶牛 繁殖性状 基因组选择 准确性 无偏性 

分 类 号：S823.2[农业科学—畜牧学]