Department of Animal Science, Iowa State University, Ames 50011, USA.
J Anim Sci. 2011 May;89(5):1270-80. doi: 10.2527/jas.2010-3107.
A 5-generation selection experiment in Yorkshire pigs for feed efficiency consists of a line selected for low residual feed intake (LRFI) and a random control line (CTRL). The objectives of this study were to use random regression models to estimate genetic parameters for daily feed intake (DFI), BW, backfat (BF), and loin muscle area (LMA) along the growth trajectory and to evaluate the effect of LRFI selection on genetic curves for DFI and BW. An additional objective was to compare random regression models using polynomials (RRP) and spline functions (RRS). Data from approximately 3 to 8 mo of age on 586 boars and 495 gilts across 5 generations were used. The average number of measurements was 85, 14, 5, and 5 for DFI, BW, BF, and LMA. The RRP models for these 4 traits were fitted with pen × on-test group as a fixed effect, second-order Legendre polynomials of age as fixed curves for each generation, and random curves for additive genetic and permanent environmental effects. Different residual variances were used for the first and second halves of the test period. The RRS models were fitted with the same fixed effects and residual variance structure as the RRP models and included genetic and permanent environmental random effects for both splines and linear Legendre polynomials of age. The RRP model was used for further analysis because the RRS model had erratic estimates of phenotypic variance and heritability, despite having a smaller Bayesian information criterion than the RRP model. From 91 to 210 d of age, estimates of heritability from the RRP model ranged from 0.10 to 0.37 for boars and 0.14 to 0.26 for gilts for DFI, from 0.39 to 0.58 for boars and 0.55 to 0.61 for gilts for BW, from 0.48 to 0.61 for boars and 0.61 to 0.79 for gilts for BF, and from 0.46 to 0.55 for boars and 0.63 to 0.81 for gilts for LMA. In generation 5, LRFI pigs had lower average genetic curves than CTRL pigs for DFI and BW, especially toward the end of the test period; estimated line differences (CTRL-LRFI) for DFI were 0.04 kg/d for boars and 0.12 kg/d for gilts at 105 d and 0.20 kg/d for boars and 0.24 kg/d for gilts at 195 d. Line differences for BW were 0.17 kg for boars and 0.69 kg for gilts at 105 d and 3.49 kg for boars and 8.96 kg for gilts at 195 d. In conclusion, selection for LRFI has resulted in a lower feed intake curve and a lower BW curve toward maturity.
一个在约克夏猪中进行的 5 世代饲料效率选择实验包括一个选择低残余采食量（LRFI）的品系和一个随机对照品系（CTRL）。本研究的目的是使用随机回归模型来估计日采食量（DFI）、BW、背膘（BF）和腰肌肉面积（LMA）在生长轨迹上的遗传参数，并评估 LRFI 选择对 DFI 和 BW 遗传曲线的影响。另一个目的是比较使用多项式（RRP）和样条函数（RRS）的随机回归模型。使用 5 个世代的 586 头公猪和 495 头母猪约 3 至 8 月龄的数据。DFI、BW、BF 和 LMA 的平均测量次数分别为 85、14、5 和 5。这些 4 个性状的 RRP 模型拟合了猪圈×试验组作为固定效应、每一代的二阶勒让德多项式固定曲线和加性遗传和永久环境效应的随机曲线。测试期的前半段和后半段使用不同的剩余方差。RRS 模型与 RRP 模型具有相同的固定效应和剩余方差结构，包括遗传和永久环境的样条和年龄的线性勒让德多项式的随机效应。选择 RRP 模型是因为尽管 RRS 模型的贝叶斯信息准则比 RRP 模型小，但它对表型方差和遗传力的估计却不稳定。在 91 至 210 日龄时，RRP 模型的遗传力估计值从公猪的 0.10 到 0.37 和母猪的 0.14 到 0.26 不等，公猪的 0.39 到 0.58 和母猪的 0.55 到 0.61 不等，公猪的 0.48 到 0.61 和母猪的 0.61 到 0.79 不等，公猪的 0.46 到 0.55 和母猪的 0.63 到 0.81 不等。在第 5 代中，LRFI 猪的 DFI 和 BW 的遗传曲线平均值低于 CTRL 猪，尤其是在测试期末；105 日龄时，DFI 的估计线差（CTRL-LRFI）为公猪 0.04kg/d，母猪 0.12kg/d，195 日龄时，公猪 0.20kg/d，母猪 0.24kg/d；105 日龄时，BW 的线差为公猪 0.17kg，母猪 0.69kg，195 日龄时，公猪 3.49kg，母猪 8.96kg。总之，LRFI 的选择导致饲料摄入量曲线和成熟时的 BW 曲线降低。
