Animal Improvement Programs Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705-2350, USA.
J Dairy Sci. 2012 Jun;95(6):3444-7. doi: 10.3168/jds.2011-5000.
Traditional evaluations of cows with genotypes have been adjusted since April 2010 to be comparable with evaluations of bulls so that their value for estimation of single nucleotide polymorphism effects in genomic evaluation programs would be improved. However, that adjustment made them not comparable with traditional evaluations of nongenotyped cows. To create an adjustment for all cows with an evaluation based on US data, Mendelian sampling, which is the difference between predicted transmitting ability (PTA) and parent average (PA), was calculated for milk, fat, and protein yields and divided by a deregression factor. Standard deviations for the deregressed Mendelian sampling (DMS) were grouped by reliability with PA contribution removed (REL(no PA)). A multiplicative adjustment to reduce the DMS standard deviation for cows so that it would be the same as for bulls with similar REL(no PA) was represented as a linear function of REL(no PA). Mean cow PA by birth year was subtracted from individual bull and cow PA to create within-year PA deviation groups, and mean DMS was calculated by PA deviation group. Means decreased for bulls and increased for cows with increasing deviation. The differences were fit by linear regression on PA deviation and used to adjust cow DMS. The adjustment reduced PTA of cows with a high PA and increased PTA of cows with a low PA but did not change estimated genetic trend because adjustment was within birth year. The adjustment also reduced variance of cow evaluations within birth year. Traditional evaluations of genotyped cows with a REL(no PA) of ≥55% were further adjusted so that the difference between those evaluations and direct genomic values calculated using only bulls as predictors was similar to that for bulls. The second adjustment was small compared with a 2010 adjustment and, therefore, had little effect on the comparability of evaluations for genotyped and nongenotyped cows. Cows with converted evaluations from other countries were excluded from the predictor population, and their converted evaluations were adjusted so that the difference between their mean PTA and direct genomic value was the same as the corresponding difference for bulls. For cows with converted evaluations, the adjustment amount differed depending on REL(no PA) (<55% or ≥55%). The new adjustment was implemented by USDA in April 2011 and permits a fairer comparison of estimated genetic merit between nongenotyped and genotyped cows.
自 2010 年 4 月以来，对具有基因型的奶牛的传统评估已进行调整，以便与公牛的评估相媲美，从而提高其在基因组评估计划中单核苷酸多态性效应估计的价值。然而，这种调整使得它们与未经基因分型的奶牛的传统评估不再可比。为了为所有基于美国数据进行评估的奶牛创建调整，计算了预测传递能力（PTA）与亲本平均值（PA）之间的 Mendelian 采样差异，并除以去回归因子。去回归 Mendelian 采样（DMS）的标准差按可靠性分组，去除 PA 贡献（REL(no PA)）。一种减少 DMS 标准差的乘法调整，以使它与具有相似 REL(no PA)的公牛相同，以 REL(no PA)的线性函数表示。通过出生年份减去个体公牛和奶牛的平均 PA 来创建年内 PA 偏差组，并按 PA 偏差组计算平均 DMS。随着偏差的增加，公牛的均值降低，奶牛的均值增加。差异通过线性回归拟合 PA 偏差，并用于调整奶牛 DMS。调整降低了高 PA 奶牛的 PTA，并增加了低 PA 奶牛的 PTA，但由于调整在出生年内，所以不会改变估计的遗传趋势。调整还降低了出生年内奶牛评估的方差。对于 REL(no PA)≥55%的基因分型奶牛的传统评估进行了进一步调整，以便这些评估与仅使用公牛作为预测因子计算的直接基因组值之间的差异与公牛相似。第二次调整与 2010 年的调整相比很小，因此对基因分型和非基因分型奶牛评估的可比性影响很小。从预测人群中排除了来自其他国家的转换评估的奶牛，并调整了他们的转换评估，以便他们的平均 PTA 和直接基因组值之间的差异与公牛的相应差异相同。对于具有转换评估的奶牛，调整量取决于 REL(no PA)（<55%或≥55%）。美国农业部于 2011 年 4 月实施了新的调整，允许在非基因分型和基因分型奶牛之间更公平地比较估计的遗传优势。
