To date, no study has tested the inclusion of bioclimatic variables or heat stress indicators in the formation of fixed regression for the genetic evaluation of dairy cattle. The objective of the present study was to investigate the inclusion of heat stress level in modeling of fixed regression in random regression models to predict population response and individual genetic components for fat and protein yield in test-day, using data from Holstein cattle in tropical environment. The data comprised 52,012 test-day fat and protein records of 9,858 first-parity Holstein cows from Brazil, collected from 1997 to 2013, and bioclimatic data (temperature-humidity index – THI, and diurnal temperature variation – DTV) from 18 weather stations. Least square linear regression models were used to determine THI and DTV thresholds for fat and protein yield losses caused by heat stress. In addition to the standard model (SM, without bioclimatic variables), THI and DTV were analyzed as fixed effects (considering the average of two days before the test-day control), and heat stress level was considered in formation of fixed regression, totaling three models for each trait. THI and DTV thresholds for fat and protein losses was THI = 74 (−0.030 kg/day/ THI) and DTV = 16 (−0.020 kg/day/ DTV), for both traits. The model that included THI and DTV as fixed effects, and used heat stress level in formation of fixed regression (fixed curve), presented a better fit (Akaike's information criterion, Schwarz's Bayesian information criterion, and residual variance). Estimated breeding values (EBV) are improved when using this model, and there is an increase in reliability of estimates. There is an important reranking of sires when heat stress indicators are included in the model to evaluate fat and protein yield in test-day. The increase in reliability of EBV by including heat stress level in the fixed regression shows a better fit of model to the data. It is possible to conclude that the inclusion of heat stress level in the formation of fixed regression is an important factor to be considered in dairy cattle reared in tropical regions to minimize environmental effects and optimize the process of evaluation and genetic selection.

迄今为止，还没有研究测试将生物气候变量或热应激指标纳入奶牛遗传评估的固定回归的形成。本研究的目的是利用热带环境中荷斯坦牛的数据，研究将热应激水平纳入随机回归模型中的固定回归建模，以预测测试日脂肪和蛋白质产量的种群反应和个体遗传成分。 . 数据包括从 1997 年到 2013 年收集的来自巴西的 9,858 头头胎荷斯坦奶牛的 52,012 测试日脂肪和蛋白质记录，以及来自 18 个气象站的生物气候数据（温湿度指数 - THI 和昼夜温度变化 - DTV）。最小二乘线性回归模型用于确定热应激引起的脂肪和蛋白质产量损失的 THI 和 DTV 阈值。除标准模型（SM，无生物气候变量）外，将THI和DTV作为固定效应进行分析（考虑测试日控制前两天的平均值），并在形成固定回归时考虑热应激水平，总计每个性状的三个模型。对于这两个性状，脂肪和蛋白质损失的 THI 和 DTV 阈值是 THI = 74（-0.030 公斤/天/ THI）和 DTV = 16（-0.020 公斤/天/ DTV）。将THI和DTV作为固定效应，利用热应力水平形成固定回归（固定曲线）的模型，拟合效果较好（赤池信息准则、施瓦茨贝叶斯信息准则、残差方差）。使用该模型时，估计育种值 (EBV) 得到改进，估计的可靠性也有所提高。当模型中包含热应激指标以评估测试日的脂肪和蛋白质产量时，对公牛进行了重要的重新排序。通过在固定回归中包含热应力水平来提高 EBV 的可靠性，这表明模型与数据的拟合更好。可以得出结论，将热应激水平纳入固定回归的形成是在热带地区饲养的奶牛中需要考虑的重要因素，以尽量减少环境影响并优化评估和遗传选择过程。当模型中包含热应激指标以评估测试日的脂肪和蛋白质产量时，对公牛进行了重要的重新排序。通过在固定回归中包含热应力水平来提高 EBV 的可靠性，这表明模型与数据的拟合效果更好。可以得出结论，将热应激水平纳入固定回归的形成是在热带地区饲养的奶牛中需要考虑的重要因素，以尽量减少环境影响并优化评估和遗传选择过程。当模型中包含热应激指标以评估测试日的脂肪和蛋白质产量时，对公牛进行了重要的重新排序。通过在固定回归中包含热应力水平来提高 EBV 的可靠性，这表明模型与数据的拟合效果更好。可以得出结论，将热应激水平纳入固定回归的形成是在热带地区饲养的奶牛中需要考虑的重要因素，以尽量减少环境影响并优化评估和遗传选择过程。