Toghiani Sajjad, VanRaden Paul M, Null Daniel J, Miles Asha M, Van Tassell Curtis P
USDA, Agricultural Research Service, Animal Genomics and Improvement Laboratory, Beltsville, MD 20705.
USDA, Agricultural Research Service, Animal Genomics and Improvement Laboratory, Beltsville, MD 20705.
J Dairy Sci. 2024 Dec;107(12):11117-11126. doi: 10.3168/jds.2024-25267. Epub 2024 Sep 27.
Most genotypes in the National Cooperator Database now originate from cows, but most previous studies validating genomic predictions have primarily focused on bulls. This study paired official within-breed genomic PTA (GPTA) and parent average (PA) for genotyped heifer calves born between 2019 and 2021 using the August 2021 database with their corresponding performance deviations (PDEV) for 17 different traits. The PDEV data became available when the heifers completed their first lactation and were extracted from the August 2023 database in which at least one PDEV value for those 17 traits existed for each genotyped heifer record. The separate breed analyses included records for 219 Ayrshires (AY), 2,715 Brown Swiss (BS), 1,055 Guernseys (GU), 949,904 Holsteins (HO), and 125,275 Jerseys (JE). These validation cows were heifer calves born between 2019 and 2021. However, due to timing or recording patterns, each trait had missing or incomplete PDEV data, leading to unbalanced distributions of records across traits. The squared accuracy of genomic prediction, or genomic reliability (r), was divided by the corresponding heritability for each trait, as only the heritable portion of cow records could be predicted, and this reliability varied across different traits and breeds. For HO and JE, the predictive ability of GPTA outperformed PA in predicting cow PDEV for yield, productive life, SCS, fertility, and health traits. The improvement ranged from 33% to 142% compared with the predictive ability of the PA. However, the results for AY, BS, and GU breeds were less consistent due to the smaller number of genotyped heifers. The r gains in those breeds were smaller and aligned with the published reliabilities of GPTA. Weighted and unweighted regressions of PDEV on GPTA and PA traits mostly exceeded the expected value of 2.00 when predicting the future trait PDEV using GPTA or PA. The larger number of observations and lower SE of the weighted regression coefficient prediction in HO and JE breeds contributed to more stable and consistent regression coefficients for all traits except milk fever and heifer livability. Our study suggests that herd owners may experience greater benefits from genomics than originally forecast.
国家合作数据库中的大多数基因型目前都来自奶牛，但之前验证基因组预测的大多数研究主要集中在公牛身上。本研究使用2021年8月的数据库，将2019年至2021年出生的基因分型小母牛犊的官方品种内基因组预测传递能力（GPTA）和父母平均值（PA）与其17种不同性状的相应性能偏差（PDEV）进行配对。当小母牛完成第一次泌乳时，PDEV数据可用，并从2023年8月的数据库中提取，其中每个基因分型小母牛记录至少存在这17种性状的一个PDEV值。单独的品种分析包括219头艾尔夏牛（AY）、2715头瑞士褐牛（BS）、1055头娟姗牛（GU）、949904头荷斯坦牛（HO）和125275头泽西牛（JE）的记录。这些用于验证的奶牛是2019年至2021年出生的小母牛犊。然而，由于时间安排或记录模式，每个性状都有缺失或不完整的PDEV数据，导致各性状的记录分布不均衡。基因组预测的平方准确率，即基因组可靠性（r），除以每个性状的相应遗传力，因为只有奶牛记录的可遗传部分可以被预测，并且这种可靠性在不同性状和品种之间有所不同。对于荷斯坦牛和泽西牛，在预测奶牛的产奶量、生产寿命、体细胞评分、繁殖力和健康性状的PDEV时，GPTA的预测能力优于PA。与PA的预测能力相比，提高幅度在33%至142%之间。然而，由于基因分型小母牛数量较少，艾尔夏牛、瑞士褐牛和娟姗牛品种的结果不太一致。这些品种的r增益较小，与已公布的GPTA可靠性一致。当使用GPTA或PA预测未来性状PDEV时，PDEV对GPTA和PA性状的加权和非加权回归大多超过预期值2.00。荷斯坦牛和泽西牛品种中观察值数量较多以及加权回归系数预测的标准误较低，使得除产乳热和小母牛存活率外的所有性状的回归系数更加稳定和一致。我们的研究表明，畜群所有者可能从基因组学中获得比最初预测更大的收益。
