ArticlePDF Available

Using Random Regression Models to Estimate Genetic Parameters on Body Weights in Layers(In chinese)

Authors:
  • Jiangsu institute of poultry science

Abstract and Figures

This study was to assess the effect of the orders of Legendre polynominals on the size of the maximum likelihood and the error, to optimize the random regression model, to evaluate the genetic potential and selection knot of layer resource population and to provide parameters for optimal layer breeding scheme for resource population. The data set consisted of 26 532 items collected from the resource population, which set up by White leghorn reciprocal crossing with the blue eggshell chickens. The pedigree consisted of 5 871 individuals. The standard of data cleaning included: i. removing outlier; ii. eliminate repeated individuals; iii. get rid of unknown sexed individuals; iv. individuals with less than 4 records were also excluded. After data cleaning procedures, 25 483records on body weight could be used in the next step. Considering the influence of the nongenetic factors on body weights, fixed effects of animal model included batch and sex factors. Using the random regression model, estimates on variance components, genetic parameters and eigenvectors were obtained. Comparing with AIC and BIC values, the best model should embed 2nd Legendre polynomials into fixed effects, 5th Legendre polynomials into additive genetic effects and permanent environmental effects. Heterogeneous residual variance was grouped into 5 levels. For the body weights on resource population, heritability ranged from 0.46 to 0.63, repeatability varied from 0.88 to 0.92, the genetic correlations ranged from 0.32 to 0.99, permanent environmental correlations varied from 0.34 to 0.99. The genetic correlations among the weeks reduced with the intervals increased, high correlations occurred between the neighboring weeks. The genetic variance, permanent environmental variance and residual variance increased with ages. The first and second eigenvalues of additive genetic effects could explain 97% of total variations. The genetic parameters on the early body weights in laying chickens were estimated with a random regression model. The individual growth curve could be altered by selection on the coefficients associated with the second eigenfunction. However, the genetic gain would be low.
Content may be subject to copyright.
中国农业科学 2020,53(11):2297-2304
Scientia Agricultura Sinica doi: 10.3864/j.issn.0578-1752.2020.11.015
收稿日期:2019-05-19接受日期:2019-10-28
基金项目:江苏现代农业产业技术体系建设项目(JATS[2018]247、现代农业产业技术体系建设专项资金CARS-40-K01、江苏省农业重大新品种
创制项目(PZCZ201729
联系方式:郭军,Tel 0514-85599012E-mailguojun.yz@gmail.com。通信作者王克华,E-mailsqbreeding@126.com
开放科学(资源服务)标识码(OSID):
应用随机回归模型估计蛋鸡体重遗传参数
郭军,曲亮,窦套存,王星果,沈曼曼,胡玉萍,王克华
(中国农业科学院家禽研究所,江苏扬州 225125)
摘要:【目的】通过分析勒让德多项式阶数对最大似然值、残差的影响,优化随机回归模型,评估蛋鸡资源
群体体重遗传潜能和选择时机,为蛋鸡资源群体育种方案提供参数。【方法】收集东乡绿壳蛋鸡与白莱航鸡 F2资源
群体体重数据 26 532 条。系谱数据包含 5 871 只鸡,其中 4 174 只鸡有 5 条记录,802 只鸡有 4 条记录,128
鸡没有记录。数据清洗包括去除离群值数据、去除翅号重复个体、去除性别不明个体、去除少于 4 条记录个体。
经整理,剩余 25 483 条体重数据,其中绿壳蛋鸡 2 223 条,白莱航鸡 696 条,F1 6 002 条,F2 16 562 条。
应用 SPSS 软件中一般线性模型分析非加性遗传因素对体重的影响,确定将批次、性别列入动物模型固定效应。应
用随机回归模型分析蛋鸡早期体重方差组分、遗传参数、随机回归系数矩阵特征向量。随机回归动物模型中包括
一般固定效应、固定回归项及随机回归项三类效应。研究中,以批次-性别作为固定效应,以周龄体重作为固定回
归项,将加性遗传效应和永久环境效应作为随机回归项。经 AIC、BIC 筛选,随机回归模型中加性遗传效应宜嵌入
5 阶勒让德多项式、永久环境效应宜嵌入 5 阶勒让德多项式、固定回归项宜嵌入 2 阶勒让德多项式。残差做异质
化处理,分为 5 个水平,即每次观测设定一个残差初始值,观测间隔期残差以线性回归计算。【结果】蛋鸡资源群
体 1—9 周龄体重遗传力为 0.46—0.63,重复力为 0.88—0.92,遗传相关系数为 0.32—0.99,永久环境相关系数
为 0.34—0.99。遗传相关系数随着周龄间隔增大而减小,相邻周龄遗传相关系数较高。遗传方差、永久环境方差
以及残差随年龄增加而增加。加性遗传效应随机回归系数矩阵前三个特征值依次为 1 976.91、161.95、42.22,
前三个特征值合计解释 99%遗传变异。【结论】随机回归模型可用于蛋鸡早期体重遗传评估及选育。对加性遗传系
数矩阵第二特征方程系数进行选择可以改变个体生长曲线,选择时机宜在 3—6 周龄。蛋鸡资源群体早期体重遗传
力略高于其它群体同类研究结果。
关键词:体重;随机回归模型;遗传力;特征值;蛋鸡
Using Random Regression Models to Estimate Genetic Parameters
on Body Weights in Layers
GUO Jun, QU Liang, DOU TaoCun, WANG XingGuo, SHEN ManMan, HU YuPing, WANG KeHua
(Poultry Institute, Chinese Academy of Agricultural Sciences, Yangzhou 225125, Jiangsu)
Abstract: Objective This study was to assess the effect of the orders of Legendre polynominals on the size of the maximum
likelihood and the error, to optimize the random regression model, to evaluate the genetic potential and selection knot of layer
resource population and to provide parameters for optimal layer breeding scheme for resource population. MethodThe data set
consisted of 26 532 items collected from the layer resource population, which set up by White Leghorn reciprocal crossing with the
blue eggshell chickens. The pedigree consisted of 5 871 individuals, including 4 147 chickens with 5 records, 802 chickens with 4
2298 53
records and 128 chickens without records. The standard of data cleaning included: i. removing outlier; ii. eliminating repeated
individuals; iii. getting rid of unknown sexed individuals; iv. individuals with less than 4 records were also excluded. After data
cleaning procedures, 25 483 records on body weight could be used in the next step, 2 223 of which collected from blue shelled
chickens, 696 of which collected from White Leghorn, 6 002 of which collected from F1 generation and 16 562 of which collected
from F2 generation. The influence of the nongenetic factors on body weights was analyzed by GLM in SPSS. The fixed effects of
animal model included batch and sex factors. Using the random regression model, variance components, genetic parameters and
eigenvectors were obtained. The model included general fixed effect and fixed regression, random regression. In this study, batch-sex
was the fixed effects, and a fixed regression was fitted for week age body weight effects; the direct additive genetic, permanent
environment were the random effects. Comparing with AIC and BIC values, the best model should embed 2nd Legendre polynomials
into fixed effects, 5th Legendre polynomials into additive genetic effects and permanent environmental effects. Heterogeneous
residual variance was grouped into 5 levels. Each observation was set an initial estimate. The residual variance between the
neighboring observations was treated as a linear regression.Result For the body weights on resource population, heritability was
ranged from 0.46 to 0.63, repeatability varied from 0.88 to 0.92, the genetic correlation was ranged from 0.32 to 0.99, and permanent
environmental correlation was varied from 0.34 to 0.99. The genetic correlations among the weeks reduced with the intervals
increased, high correlations occurred between the neighboring weeks. The genetic variance, permanent environmental variance and
residual variance increased with ages. The first three eigenvalues of additive genetic effects was 1 976.91, 161.95, and 42.22,
respectively, and these eigenvalues could explain 99% of total variations. ConclusionThe genetic parameters on the early body
weights in laying chickens were estimated with a random regression model. The individual growth curve could be altered by
selection on the coefficients associated with the second eigenfunction. The right time seemed to select on 3 to 6 week. Estimates of
heritability in the resource population were larger than the results in the literatures.
Key words: body weight; random regression model; heritability; eigenvalue; layer
0 引言
【研究意义】体重是蛋鸡选育目标之一。为节约
生产成本、减少饲料消耗,蛋鸡体重应在满足产蛋性
能需要的前提下尽量减少[1]ANDERSON 等比较分析
19582011 年北卡随机交配群体与商品蛋鸡群体
生产性能评测试验,结果表明商品蛋鸡早期体重持续
降低[2]。绿壳蛋鸡已在我国商品化生产,然而绿壳蛋
鸡体重遗传力等参数尚未见报到,影响了绿壳蛋鸡选
择准确性。【前人研究进展】蛋鸡体重可以重复测量,
并随测量时间呈现连续递进变化,此类性状称为纵向
性状,也称动态性状[3]、函数值性状[4]。以随机回归
模型解析动态性状遗传参数是当前流行趋势。近年
来,利用随机回归模型评估家禽体重研究取得一些进
展。BEGLI 等以艾维因肉鸡与伊朗地方鸡 F2资源群体
为素材,评估体重与饲料利用效率性状[5]。同样针对
肉鸡,MEBRATIE 等应用随机回归模型分析了科宝肉
鸡体重性状,结果显示遗传背景随生长阶段推移而改
[6]ROVADOSCKI [7]以随机回归模型分析了散养
条件下巴西 4个试验品系体重遗传参数。RAFAT [8]
应用随机回归模型解析了火鸡 232 周龄体重遗传
力、遗传相关系数,发现随机回归模型不需要校正数
据,还比传统模型提供了更多信息。除了鸡和火鸡之
外,应用随机模型还评估了肉用鹌鹑和蛋用鹌鹑体重
性状[9-12]。【本研究切入点】综上所述,多个团队以
随机回归模型评估了肉用家禽体重性状。针对蛋鸡尚
未见应用随机回归模型评估体重性状。【拟解决的关
键问题】本研究以白莱航鸡与东乡绿壳蛋鸡 F2资源群
体为素材,收集 20102013 年体重数据,剖分方差组
分,解析遗传参数,分析特征值及特征方程。研究结
果将为绿壳蛋鸡选育提供支持。
1 材料与方法
1.1 试验动物
体重数据采集自江苏省家禽研究所邵伯基地(江
苏省扬州市江都区邵伯镇小街 2号)F2蛋鸡资源群体,
有关群体构建信息详见文献[13]。简言之,以东乡绿
壳蛋鸡、白莱航鸡为亲本,经正反交获得 F1代、F2
代,其中 F1代出雏 1 581 只鸡,F2代出雏 3 749 只鸡。
亲代出雏时间为 2011 821 日,F1出雏时间为 2012
75日,F2出雏时间为 2013 25日。试验
鸡出雏时戴翅号,翻肛鉴别雌雄。育雏 19周龄期间,
隔周称重,称重前禁食 12 h
1.2 统计分析
1.2.1 数据整理 蛋鸡资源群体体重原始数据共有
26 532 条记录,其中 1周龄 5 303 条、3周龄 5 280 条、
11 郭军等:应用随机回归模型估计蛋鸡体重遗传参数 2299
5周龄 5 310 条、7周龄 5 275 条、9周龄 5 353 条。数
据清洗包括去除翅号重复个体、去除离群值、去除性别
不明个体、去除少于 4条记录个体。经数据清洗后,
资源群体体重数据集剩余 25 483 条记录。其中绿壳蛋
2 223 条,白莱航鸡 696 条,F16 002 条,F2
16 562 条。系谱数据包含 5 871 只鸡,其中 4 174 只鸡
5条记录,802 只鸡有 4条记录,128 只鸡没有记录。
ANOVA 分析批次、性别、母亲体重对蛋鸡体重的
影响,确定影响体重的非加性遗传因素。1列出资源
群体体重表型值信息。
表1 蛋鸡资源群体体重数据统计
Table 1 Statistics on body weights in the resource population
世代
Generation
1周龄体重
1st-week BW
3周龄体重
3rd-week BW
5周龄体重
5th-week BW
7周龄体重
7th-week BW
9周龄体重
9th-week BW
绿壳蛋鸡公鸡 Blue shelled cocks 48.53b±5.24 107.40b±14.51 167.13b±27.25 301.11b±47.41 469.60b,c±68.73
绿壳蛋鸡母鸡 Blue shelled hens 46.29 a±5.14 97.05a±12.56 147.23a±24.11 260.84a±37.50 396.74a±52.44
白莱航鸡公鸡 White leghorn cocks 69.54f±6.80 162.88g±14.94 268.08f±28.13 473.17f±38.15 634.12g±67.65
白莱航鸡母鸡 White leghorn hens 66.02e±6.75 145.29e±13.31 236.76d±23.36 397.29e±36.25 539.40e±55.33
F1代公鸡 Cocks in F1 generation 47.05a±7.74 141.15d±20.34 273.07f±31.35 382.72e±48.58 569.77f±67.75
F1代母鸡 Hens in F1 generation 45.64a±7.35 130.03c±19.90 241.94d±29.27 331.45c±39.61 478.98c±58.73
F2代公鸡 Cocks in F2 generation 62.91d±5.82 152.26f±19.45 252.52e±34.15 393.87e±56.56 522.68d±71.82
F2代母鸡 Hens in F2 generation 60.61c±5.33 139.27d±15.96 225.19c±28.33 353.03d±47.51 458.44b±56.63
体重平均值标为 SNK 两两比较结果,不同字母表示差异显著(P0.05
Different letters in same column indicate significant difference among groups as determined by SNK multiple comparisons(P0.05)
1.2.2 遗传模型 WOMBAT 软件估计方差组分、
遗传参数,并预测育种值[14]。残差处理影响遗传评估
结果准确性。本研究以不同水平处理残差,即异质残
差方法。首先,以单性状动物模型按周龄分别分析体
重数据,残差相近的划为同一水平。依据单性状模型
分析结果,确定残差水平以及初始设定值。然后,以
随机回归模型评估资源群体体重性状,当参数向量变
化小于 10-8、对数似然值变化小于 5×10-4 时,运行达
到收敛标准,完成计算。随机回归模型数学表达式为,
3
12
** *
00 0
() () ()
n
nn
ikl l m km m km m ikl
mm m
yFE bt a t p te
φφφ
== =
=+ + + +
∑∑ ∑
式中,yikl 是第 i批次第 l周龄第 k只鸡体重;FE 是固
定效应,包括性别效应和世代效应;bl是第 l周龄第 m
个固定回归系数;akm 是第 k只鸡加性效应第 m个随
机回归系数;pkm 是第 k只鸡永久环境效应第 m个随
机回归系数;t*是标准化的时间,取值在-1 +1 之间;
φm是嵌入的勒让德多项式;eikl 是残差效应;n1n2
n3是嵌入固定效应、加性遗传以及永久环境效应勒让
德多项式阶数。
1.2.3 方差参数及特征方程 计算遗传参数过程包
括方差组分剖分、标准误计算以及特征方程分析。遗
传、永久环境方差计算公式为,
2'
aa
K
σ
φφ
=
2'
p
epe
K
σ
φφ
=
式中, 2
a
σ
2
p
e
σ
为蛋鸡体重加性遗传、永久环境方差。
φ
为勒让德多项式回归系数矩阵,由标准化的时间与
勒让德多项式相乘。KaKpe 为随机回归系数矩阵,
Wombat 软件获得。
遗传相关标准误计算公式为,
22
2
22
() ()
1
2
Axy
x
y
A
r
xy
SE h SE h
r
SE hh
=
式中,SE 为标准误,rAxy日龄遗传相关,h2
遗传力,xy分别为日龄。
K矩阵的特征值及特征向量通过 R计算获得,
计算代码为 eigen(K).为获得特征方程,需计算 E
E
为特征向量,
Λ
为勒让德多项式系数矩阵[15]5阶多
项式系数矩阵为,
5
0.7071 0 0.7906 0 0.7950
0 1.2247 0 2.8063 0
0 0 2.3717 0 7.9550
0 0 0 4.6771 0
0 0 0 0 9.2808
Λ=
1.2.4 模型比较 使用赤池弘次信息准则Akaike’s
information criterionAIC)以及贝叶斯信息准则
Bayesian Information CriterionBIC选择最优模型。
2300 53
AICBIC 分别写作,
AIC=-2lgL+2p
BIC=-2lgL+plg(N-r)
其中,L为模型最大似然值的对数,p为模型参数数量,
N为观察值总数,r为固定效应指示矩阵阶数。
AICBIC 都对过多参数进行惩罚,相对而言
BIC 惩罚程度更大。优先选择 AICBIC 值偏小的
模型[16]
2 结果
2.1 固定因子确定
由表 1可知,蛋鸡体重标准差随周龄增加而增加。
同周龄体重 SNK 多重比较可知,白莱航公鸡较重,绿
壳蛋鸡母鸡较轻。经 SPSS 直方图、P-P 图及 Q-Q
分析体重数据符合正态分布。ANOVA 分析表明,公
鸡体重高于同周龄母鸡体重(P0.001),同周龄世
代之间存在显著差异P0.05因此,动物模型将
批次(世代)和性别列入固定效应。
2.2 模型选择
依据表 2列出的 AICBIC 选择标准,遗传模型
加性遗传效应宜包含 5阶勒让德多项式,永久环境模
型宜包含 5阶勒让德多项式,固定效应宜包含 2阶勒
让德多项式。设定异质性残差是由于生理条件、健康
状态以及一些未知因素导致的不同周龄体重上的差
异。残差设定为 5个水平,即每次观测设定一个残差
初始值,观测间隔期残差以线性回归计算。
2.3 遗传参数估计
随机回归模型以无限维形式呈现方差-协方差矩
阵,表 3按周龄给出体重方差组分及遗传参数。加性
遗传方差、永久环境方差及残差演变趋势基本一致,
即随周龄增加而增加。加性遗传方差大于同周龄永久
环境方差。加性遗传方差和永久环境方差标准误都很
小,大部分变异系数在 10%以内,表明方差参数估值
比较精确。
3列出蛋鸡资源群体 19周龄体重遗传力和重
复力。绿壳蛋鸡与白莱航鸡资源群体体重遗传力为
0.460.63最高值出现在第 2周。除第 2周之外,
传力走势趋向两头低中间高的钟型。蛋鸡资源群体早
期体重遗传力为中等偏高,表明该群体体重性状具有
较多遗传潜能。永久环境效应占比与遗传力走势不同,
呈现两头高中间低的形势。蛋鸡体重重复力维持在
0.90 左右,表明蛋鸡体重比较稳定,还表明重复测量
差异主要来自个体间差异。
各周龄体重遗传相关系数、永久环境相关系数见
4,相邻周龄遗传相关系数和永久环境相关系数较
高,两种相关系数随着周龄间隔增大而减小。各周龄
永久环境效应为中等到高等相关,表明永久环境对体
重有重要影响。
2 蛋鸡体重遗传模型比较
Table 2 Comparison of models on body weight in laying chickens
加性遗传效应阶数
Orders of additive genetic effects
永久环境效应阶数
Orders of PE effects
参数个数
The number of parameters
最大对数似然值
Maximum lgL
AIC BIC
2 2 11 -84372 168766 168855
2 3 14 -83807 167642 167755
2 4 18 -83776 167589 167734
2 5 23 -83734 167514 167514
2 6 29 -83734 167526 167760
3 3 17 -83627 167288 167426
3 4 21 -83599 167240 167410
3 5 26 -83553 167157 167367
3 6 32 -83553 167169 167428
4 4 25 -83515 167081 167283
4 5 30 -83419 166899 167142
4 6 36 -83419 166911 167202
5 5 35 -83052 166173 166456
5 6 41 -83052 166185 166517
6 6 47 -83052 166197 166577
11 郭军等:应用随机回归模型估计蛋鸡体重遗传参数 2301
2.4 特征方程分析
加性遗传随机回归系数矩阵特征值依次为:
1 976.91
161.9542.2216.720.22所占百分比为 89.947.37
1.920.760.01。第 4特征值及后面的特征值占比少
1%,对蛋鸡早期生长的贡献很少,故忽略不计。
如图 1所示,加性遗传第 1特征方程近乎线性,全程
为正值。第 2特征方程呈现余弦波走势,1946
龄以及邻近第 9周为正值,其余为负值。第 3特征
方程为波浪状,17周龄为正值,8周为负值且
呈急剧下降。永久环境随机回归系数矩阵特征值依次
为:1 341.10124.7035.511.530.57,所占百分
比为 89.208.292.360.100.04
3 讨论
传统上,畜禽遗传评估以多性状动物模型分析重
复测量育种数据。与多性状动物模型相比,随机回归
模型的优点在于:①可以利用特征值系数有效改变性
状曲线;②以较少的参数拟合方差-协方差矩阵;③能
体重变化 Body weight change
1 体重遗传关系矩阵前 3 个特征值及相应特征方程
Fig. 1 Three largest eigenvalues of the eigenfunctions on body weight
3 蛋鸡体重方差组分和遗传参数
Table 3 Variance component and genetic parameters on body weight in resource population
周龄
Weeks
加性遗传方差
Additive variance
永久环境方差
Permanent environmental variance
残差
Residual
表型方差
Phenotypic variance
遗传力
Heritability
永久环境方差占比
22
/
p
ep
σ
σ
重复力
Repeatability
1 18.91±1.44 14.76±0.85 3.99 37.67±1.01 0.50±0.03 0.39±0.03 0.891
2 158.36±13.51 70.01±8.23 22.85 251.21±9.07 0.63±0.04 0.28±0.04 0.91
3 214.60±14.83 153.93±8.80 41.70 410.23±10.50 0.52±0.03 0.38±0.02 0.90
4 538.68±32.11 349.48± 18.39 87.74 975.89±24.30 0.55±0.02 0.36±0.02 0.91
5 925.68±55.51 528.66± 30.59 133.77 1588.11±40.61 0.58±0.02 0.33±0.02 0.92
6 1275.48±86.87 694.50± 47.56 264.02 2233.99±57.81 0.57±0.03 0.31±0.03 0.88
7 1882.89±152.26 1100.43± 85.98 394.26 3377.59±95.76 0.56±0.03 0.33±0.03 0.88
8 2471.92±216.91 1798.51±126.09 498.20 4768.62±135.92 0.52±0.03 0.38±0.03 0.90
9 2797.48±238.06 2731.94±147.03 602.13 6131.55±159.58 0.46±0.03 0.45±0.03 0.90
重复力标准误接近于 0,故省略 Considering the standard error of repeatability approached zero, this item was omitted
2302 53
4 不同周龄蛋鸡体重遗传相关系数(对角线下)及永久环境相关系数(对角线上)
Table 4 Estimated values of genetic (below the diagonal) and permanent environmental correlations (above the diagonal) on body
weight
周龄 Weeks 周龄
Weeks 1 2 3 4 5 6 7 8 9
1 0.59±0.05 0.65±0.03 0.59±0.03 0.54±0.03 0.48±0.05 0.38±0.05 0.34±0.05 0.42±0.04
2 0.68±0.02 0.76±0.03 0.49±0.05 0.43±0.05 0.48±0.06 0.52±0.06 0.54±0.05 0.50±0.05
3 0.86±0.01 0.71±0.02 0.93±0.01 0.87±0.01 0.77±0.02 0.62±0.03 0.57±0.03 0.71±0.02
4 0.71±0.02 0.32±0.03 0.90±0.01 0.97±0.00 0.85±0.01 0.64±0.03 0.57±0.03 0.76±0.02
5 0.68±0.02 0.33±0.03 0.89±0.01 0.99±0.00 0.94±0.01 0.77±0.02 0.71±0.02 0.86±0.01
6 0.70±0.02 0.54±0.03 0.91±0.01 0.91±0.01 0.95±0.00 0.94±0.01 0.90±0.01 0.98±0.00
7 0.67±0.02 0.72±0.02 0.86±0.01 0.74±0.01 0.80±0.01 0.95±0.00 0.99±0.00 0.97±0.00
8 0.66±0.02 0.73±0.02 0.82±0.01 0.67±0.02 0.73±0.01 0.90±0.01 0.99±0.00 0.96±0.00
9 0.67±0.02 0.40±0.04 0.78±0.02 0.82±0.01 0.86±0.01 0.89±0.01 0.85±0.01 0.87±0.01
够利用不完整记录[15, 17]。本研究利用蛋鸡资源群体体
重动态数据,以动物模型估计遗传参数,分析早期体
重遗传潜能以及选择时机。为实现研究主旨,需要掌
控好两个关键点,即模型的优化和特征方程的建立。
特征方程可用于选育蛋鸡体重。加性遗传系数矩
阵特征值具有生物学意义,它是遗传变异在各个维度
的最大选育潜力。本研究中,随机回归模型赋予加性
遗传系数矩阵时间属性,因而通过特征向量建立以时
间为自变量的特征方程。加性遗传系数矩阵第一特征
值反应蛋鸡线性生长趋势。蛋鸡体重第一特征方程在
19周时间段内为正值,表明选择任意时间点体重也
搭车选择其他时间点,无论选择增加体重还是选择减
少体重。应用随机回归模型分析鹌鹑体重,结果表明
加性遗传效应系数矩阵第 1特征值及其特征向量约占
总变异 87%,与本研究结果相当[18]BOLIGON [16]
分析了内洛尔牛初生重至成年体重遗传参数变化,结
果表明第 1特征值约占总变异的 90%ABEGAZ [19]
以随机回归模型分析了埃塞俄比亚奥罗绵羊体重遗传
变异规律,结果表明第 1特征值约占总遗传变异 83%
与本文结果相近。然而,肉牛体重遗传评估结果表明,
1特征值解释了 96%遗传变异[20]肉牛上的研究结
果不同于其它物种,或许归因于遗传背景,也可能因
为肉牛体重遗传评估时使用了降维处理方式,即忽略
AICBIC 选择,强行将随机回归系数矩阵降阶。加
性遗传系数矩阵第二特征值可用于改变个体生长发育
曲线。然而,考虑到第 2特征值在遗传变异中的占比,
选择该系数的遗传进展将会缓慢。
MEYER [21]指出,
特征值占比高于 5%的选择都有意义,因而可以对资
源群体加性遗传系数矩阵第二特征值进行选择。本
团队应用随机回归模型还分析了蛋鸡资源群体蛋黄
[13]、蛋壳强度[22],但第二特征值占比低于 5%,表
明这两个性状不适合特征值选择。其他特征值因为总
变异占比较少,选育时可忽略。
随机回归模型多项式阶数对于遗传参数、方差组
分估计值很重要。POOL 等认为,3阶勒让德多项式
足以满足遗传评估生长或泌乳曲线的需要[23]。高阶多
项式不仅对计算能力提出更高要求,还带来标准误增
大。育种实践中,BLUP 方法容错性能较好,即使参
数存在轻微偏差,也不至于影响育种值排序。然而,
BROTHERSTONE [24]认为当数据分析难于收敛时,
高阶多项式比低阶多项式效率更高。LEGARRA [25]
认为降低多项式维数可能造成端值异常。本研究主旨
在于估计遗传参数,故遗传模型应用高阶勒让德多项
式。分析结果显示两个端值遗传力未出现异常。同一
群体的另外两个性状,蛋黄重和蛋壳强度,由于应用
低阶多项式处理,导致端值遗传力偏大[13,22]。总之,
随机回归模型的优化取决于研究目的。如果估计遗传
参数,选择高阶多项式;如果预测育种值,考虑低阶
多项式。
早期体重不仅与鸡的维持需要有关,还与个体健
康程度、均匀度存在关联,因而国内外学者针对早期
体重开展了遗传评估工作。DRUYAN [26]分析了安
卡肉鸡早期体重遗传参数,71419 日龄体重遗传
力分别为 0.560.540.63,与本研究结果相近。
MOGHADAM [27]以多性状动物模型分析了肉鸡猝
死、腹水与体重的关系,获得考尼仕鸡 4周龄体重遗
11 郭军等:应用随机回归模型估计蛋鸡体重遗传参数 2303
传力为 0.45白洛克鸡遗传力为 0.35 的研究结果,
于本文报告的结果。
GAYA [28]以多性状动物模型分
析罗斯肉鸡 38 日龄体重,获得其遗传力为 0.40,低 于
本研究结果。本研究团队以如皋黄鸡为试验材料,应用
REML 方法评估其 6周龄体重加性遗传力为 0.33,低
于本研究结果[29]。以上鸡早期体重遗传分析结果存在
差异,究其原因,与遗传背景不同有关,也与研究模型
以及固定效应设置有关。本文遗传参数估计值略高于其
它研究结果,与所用素材来自 F2杂交群体有关。
4 结论
随机回归模型可用于蛋鸡早期体重选育。研究结
果表明:
1动物模型嵌入 5阶勒让德多项式可以拟合
加性遗传效应和永久环境效应;
2遗传相关和特征方
程分析表明资源群体可是进行体重早期选择,通过选
择特征值系数可以改变个体生长曲线,选择时机宜在
36周龄;3蛋鸡资源群体体重重复力在 0.90 左右,
表明 23次测量即可准确预测今后的体重水平;4
蛋鸡资源群体早期体重遗传力略高于其他群体同类研
究结果。后续研究将进一步加强数据采集分析工作。
如果体重数据能够包括 9周龄之后记录,就能获知加
性遗传第一特征方程能否回落,第二特征方程能否持
续余弦波走势,对蛋鸡资源群体的选育工作提供更多
支持。
References
[1] GOODMAN B, GODFREY G F. Heritability of body weight in the
domestic fowl. Poultry Science, 1956, 35(1): 50-53.
[2] ANDERSON K, HAVENSTEIN G, JENKINS P, OSBORNE J.
Changes in commercial laying stock performance, 1958-2011:
Thirty-seven flocks of the North Carolina random sample and
subsequent layer performance and management tests. World's Poultry
Science Journal, 2013, 69(3): 489-514.
[3] 杨运清, 李仁杰, 李淑玲. 动态性状遗传参数的估计方法. 畜牧兽
医学报, 1996(5): 412-416.
YANG Y Q, LI R J, LI S L. An estimation method for genetic
parameters of dynamic traits. Acta Veterinaria et Zootechnica Sinica,
1996(5): 412-416.(in Chinese)
[4] JAFFRÉZIC F, PLETCHER S D. Statistical models for estimating the
genetic basis of repeated measures and other function-valued traits.
Genetics, 2000, 156(2): 913-922.
[5] BEGLI H E, TORSHIZI R V, MASOUDI A A, EHSANI A, JENSEN J.
Longitudinal analysis of body weight, feed intake and residual feed
intake in F-2 chickens. Livestock Science, 2016, 184: 28-34.
[6] MEBRATIE W, SHIRALI M, MADSEN P, SAPP R L, HAWKEN R,
JENSEN J. The effect of selection and sex on genetic parameters of
body weight at different ages in a commercial broiler chicken
population. Livestock Science, 2017, 204: 78-87.
[7] ROVADOSCKI G A, PETRINI J, RAMIREZ-DIAZ J, PERTILE S F
N, PERTILLE F, SALVIAN M, IUNG L H S, RODRIGUEZ M A P,
ZAMPAR A, GAYA L G, CARVALHO R S B, COELHO A A D,
SAVINO V J M, COUTINHO L L, MOURAO G B. Genetic
parameters for growth characteristics of free-range chickens under
univariate random regression models. Poultry Science, 2016, 95(9):
1989-1998.
[8] RAFAT S A, NAMAVAR P, SHODJA D J, JANMOHAMMADI H,
KHOSROSHAHI H Z, DAVID I. Estimates of the genetic parameters
of turkey body weight using random regression analysis. Animal, 2011,
5(11): 1699-1704.
[9] MOTA L F M, MARTINS P, LITTIERE T O, ABREU L R A, SILVA
M A,BONAFE C M. Genetic evaluation and selection response for
growth in meat-type quail through random regression models using
B-spline functions and Legendre polynomials. Animal, 2018, 12(4):
667-674.
[10 ] KARAMI K, ZEREHDARAN S, TAHMOORESPUR M, BARZANOONI
B, LOTFI E. Genetic evaluation of weekly body weight in Japanese
quail using random regression models. British Poultry Science, 2017,
58(1): 13-18.
[11] BARBIERI A, ONO R K, CURSINO L L, FARAH M M, PIRES M P,
BERTIPAGLIA T S, PIRES A V, CAVANI L, CARRENO L O
D,FONSECA R. Genetic parameters for body weight in meat quail.
Poultry Science, 2015, 94(2): 169-171.
[12] SILVA L P, RIBEIRO J C, CRISPIM A C, SILVA F G, BONAFE C M,
SILVA F F, TORRES R A. Genetic parameters of body weight and egg
traits in meat-type quail. Livestock Science, 2013, 153(1-3): 27-32.
[13] 郭军, 王克华, 曲亮, 沈曼曼, 马猛, 窦套存,胡玉萍. 应用随机回
归模型分析蛋鸡蛋黄质量遗传参数. 南京农业大学学报, 2016,
39(1): 145-149.
GUO J, WANG K H, QU L, SHEN M M, MA M, DOU T C, HU Y P.
Genetic parameters for yolk weight in resource population with
random regression model. Journal of Nanjing Agricultural University,
2016, 39(1): 145-149. (in Chinese)
[14] MEYER K. WOMBAT—A tool for mixed model analyses in
quantitative genetics by restricted maximum likelihood (REML).
Journal of Zhejiang University Science B, 2007, 8(11): 815-821.
[15] KIRKPATRICK M, LOFSVOLD D, BULMER M. Analysis of the
2304 53
inheritance, selection and evolution of growth trajectories. Genetics,
1990, 124(4): 979-993.
[16] BOLIGON A A, MERCADANTE M E Z, FORNI S, LÔBO R
B,ALBUQUERQUE L G. Covariance functions for body weight from
birth to maturity in Nellore cows. Journal of Animal Science, 2010,
88(3): 849-859.
[17] SCHAEFFER L. Application of random regression models in animal
breeding. Livestock Production Science, 2004, 86(1/3): 35-45.
[18] AKBAS Y, TAKMA C, YAYLAK E. Genetic parameters for quail
body weights using a random regression model. South African Journal
of Animal Science, 2004, 34(2): 104-109.
[19] ABEGAZ S, VAN WYK J B, OLIVIER J J. Estimates of (co) variance
function for growth to yearling in Horro sheep of Ethiopia using
random regression model. Archives Animal Breeding, 2010, 53(6):
689-700.
[20] ARANGO J, CUNDIFF L,VAN VLECK L D. Covariance functions
and random regression models for cow weight in beef cattle. Journal
of Animal Science, 2004, 82(1): 54-67.
[21] MEYER K, KIRKPATRICK M. Restricted maximum likelihood
estimation of genetic principal components and smoothed covariance
matrices. Genetics Selection Evolution, 2005, 37(1): 1.
[22] GUO J, MA M, QU L, SHEN M M, DOU T C, WANG K H.
Estimation of genetic parameters related to eggshell strength using
random regression models. British Poultry Science, 2015, 56(6):
645-650.
[23] POOL M, MEUWISSEN T. Reduction of the number of parameters
needed for a polynomial random regression test day model. Livestock
Production Science, 2000, 64(2/3): 133-145.
[24] BROTHERSTONE S, WHITE I, MEYER K. Genetic modelling of
daily milk yield using orthogonal polynomials and parametric curves.
2000, 70(3): 407-415.
[25] LEGARRA A, MISZTAL I, BERTRAND J. Constructing covariance
functions for random regression models for growth in Gelbvieh beef
cattle. Journal of Aanimal Science, 2004, 82(6): 1564-1571.
[26] DRUYAN S, SHLOSBERG A,CAHANER A. Evaluation of growth
rate, body weight, heart rate, and blood parameters as potential
indicators for selection against susceptibility to the ascites syndrome
in young broilers. Poultry Science, 2007, 86(4): 621-629.
[27] MOGHADAM H K, MCMILLAN I, CHAMBERS J R, JULIAN R J,
TRANCHANT C C. Heritability of sudden death syndrome and its
associated correlations to ascites and body weight in broilers. British
Poultry Science, 2005, 46(1): 54-57.
[28] GAYA L G, FERRAZ J B S, REZENDE F M, MOURÃO G B,
MATTOS E C, ELER J P, MICHELAN FILHO T. Heritability and
genetic correlation estimates for performance and carcass and body
composition traits in a male broiler line. Poultry Science, 2006, 85(5):
837-843.
[29] 窦套存, 王克华, 曲亮, 沈曼曼, 胡玉萍, 郭军. 如皋黄鸡 42 日龄
体重母体遗传效应评估. 山西农业大学学报(自然科学版), 2014,
34(6): 511-517.
DOU T C, WANG K H, QU L, SHEN M M, HU Y P, GUO J.
Maternal genetic effects of body weight in Rugao Yellow Chicken.
Journal of Shanxi Agricultural University(Nature Science Edition),
2014, 34(6): 511-517. (in Chinese)
(责任编辑 林鉴非)
Article
Full-text available
Background】 In addition to regulating its own phenotype, the genotype of an individual animal also affects the performance of other animals within a social group through resource allocation or behavioral interaction, and this phenomenon is called an indirect genetic effect. In the animal breeding, if the genetic model harbored the indirect genetic effects, it will not only improve the social relationship between individuals, but also obtain more genetic gains. 【Objective】In this study, Rugao Yellow Chickens raised in group cages were used as the test animals, and the indirect genetic model was used to evaluate body weight data, aiming to provide a flexible model to select the Rugao Yellow Chickens. 【Method】 The body weight data was collected from the breeding group of Rugao
Article
Full-text available
The objective was to estimate (co)variance functions using random regression models (RRM) with Legendre polynomials, B-spline function and multi-trait models aimed at evaluating genetic parameters of growth traits in meat-type quail. A database containing the complete pedigree information of 7000 meat-type quail was utilized. The models included the fixed effects of contemporary group and generation. Direct additive genetic and permanent environmental effects, considered as random, were modeled using B-spline functions considering quadratic and cubic polynomials for each individual segment, and Legendre polynomials for age. Residual variances were grouped in four age classes. Direct additive genetic and permanent environmental effects were modeled using 2 to 4 segments and were modeled by Legendre polynomial with orders of fit ranging from 2 to 4. The model with quadratic B-spline adjustment, using four segments for direct additive genetic and permanent environmental effects, was the most appropriate and parsimonious to describe the covariance structure of the data. The RRM using Legendre polynomials presented an underestimation of the residual variance. Lesser heritability estimates were observed for multi-trait models in comparison with RRM for the evaluated ages. In general, the genetic correlations between measures of BW from hatching to 35 days of age decreased as the range between the evaluated ages increased. Genetic trend for BW was positive and significant along the selection generations. The genetic response to selection for BW in the evaluated ages presented greater values for RRM compared with multi-trait models. In summary, RRM using B-spline functions with four residual variance classes and segments were the best fit for genetic evaluation of growth traits in meat-type quail. In conclusion, RRM should be considered in genetic evaluation of breeding programs.
Article
Full-text available
Repeated measures from the same individual have been analyzed by using repeatability and finite dimension models under univariate or multivariate analyses. However, in the last decade, the use of random regression models for genetic studies with longitudinal data have become more common. Thus, the aim of this research was to estimate genetic parameters for body weight of four experimental chicken lines by using univariate random regression models. Body weight data from hatching to 84 days of age (n = 34,730) from four experimental free-range chicken lines (7P, Caipirão da ESALQ, Caipirinha da ESALQ and Carijó Barbado) were used. The analysis model included the fixed effects of contemporary group (gender and rearing system), fixed regression coefficients for age at measurement, and random regression coefficients for permanent environmental effects and additive genetic effects. Heterogeneous variances for residual effects were considered, and one residual variance was assigned for each of six subclasses of age at measurement. Random regression curves were modeled by using Legendre polynomials of the second and third orders, with the best model chosen based on the Akaike Information Criterion, Bayesian Information Criterion, and restricted maximum likelihood. Multivariate analyses under the same animal mixed model were also performed for the validation of the random regression models. The Legendre polynomials of second order were better for describing the growth curves of the lines studied. Moderate to high heritabilities (h2 = 0.15 to 0.98) were estimated for body weight between one and 84 days of age, suggesting that selection for body weight at all ages can be used as a selection criteria. Genetic correlations among body weight records obtained through multivariate analyses ranged from 0.18 to 0.96, 0.12 to 0.89, 0.06 to 0.96, and 0.28 to 0.96 in 7P, Caipirão da ESALQ, Caipirinha da ESALQ, and Carijó Barbado chicken lines, respectively. Results indicate that genetic gain for body weight can be achieved by selection. Also, selection for body weight at 42 days of age can be maintained as a selection criterion.
Article
Full-text available
Random regression analyses of weight data from birth to 396 days were done using 22 141 weight records of 1 951 Horro lambs. Six different models formed from three different orthogonal polynomial regressions (legendre scale)orders (quadratic, cubic, quartic) of fit for both additive genetic and animals' permanent environmental effects, with assumption of either homogeneous or heterogeneous residual variance, were compared. Fixed effects of year and type of birth, sex and age of dam were fitted along with a fourth order polynomial. Both likelihood ratio test (LRT) and Akaike's Information Criterion (AIC) were used for model comparison. Model fit improved with increased order of polynomial and assumption of heterogeneity of residual variance. Components for additive genetic and permanent environmental (co)variance increased from 0.03 and 0.09 at birth to 23.8 and 37.6 at 396 days of age, respectively. The first three eigenvalues of the coefficient matrix of the additive genetic covariance accounted for about 98 % of the sum of all the eigenvalues. Heritability estimates have shown a declining and increasing trend at different parts of the trajectory, the lowest estimate being 0.14 for weight at birth while the highest being 0.36 for weight at about 390 days of age. Higher heritability estimates in previous uni- and bi-variate models and in the current study and also strong correlation with weight at early age makes weight at one year of age the most important trait to consider in improving productivity in Horro sheep. © Leibniz Institute for Farm Animal Biology, Dummerstorf, Germany.
Article
Full-text available
1. This study examined the changes in eggshell strength and the genetic parameters related to this trait throughout a hen's laying life using random regression. 2. The data were collected from a crossbred population between 2011 and 2014, where the eggshell strength was determined repeatedly for 2260 hens. 3. Using random regression models (RRM), several Legendre polynomials were employed to estimate the fixed, direct genetic and permanent environment effects. The residual effects were treated as independently distributed with heterogeneous variance for each test week. 4. The direct genetic variance was included with second order Legendre polynomials and the permanent environment with third order Legendre polynomials. 5 The heritability of eggshell strength ranged from 0.26 to 0.43, the repeatability ranged between 0.47 and 0.69, and the estimated genetic correlations between test weeks was high at > 0.67. 6. The first eigenvalue of the genetic covariance matrix accounted for about 97% of the sum of all the eigenvalues. 7. The flexibility and statistical power of RRM suggest that this model could be an effective method to improve eggshell quality and to reduce losses due to cracked eggs in a breeding plan.
Article
Full-text available
Abstract Principal component analysis is a widely used 'dimension reduction' technique, albeit generally at a phenotypic level. It is shown that we can estimate genetic principal components directly through a simple reparameterisation of the usual linear, mixed model. This is applicable to any analysis fitting multiple, correlated genetic effects, whether effects for individual traits or sets of random regression coefficients to model trajectories. Depending on the magnitude of genetic correlation, a subset of the principal component generally suffices to capture the bulk of genetic variation. Corresponding estimates of genetic covariance matrices are more parsimonious, have reduced rank and are smoothed, with the number of parameters required to model the dispersion structure reduced from k(k + 1)/2 to m(2k - m + 1)/2 for k effects and m principal components. Estimation of these parameters, the largest eigenvalues and pertaining eigenvectors of the genetic covariance matrix, via restricted maximum likelihood using derivatives of the likelihood, is described. It is shown that reduced rank estimation can reduce computational requirements of multivariate analyses substantially. An application to the analysis of eight traits recorded via live ultrasound scanning of beef cattle is given.
Article
Full-text available
BODY weight of broilers has been an important trait for a considerable number of years. More recently, breeders of egg production strains have become interested in adult body weight; the trend seems to be toward a small-bodied layer in order to reduce the amount of feed required for body maintenance. Heritability estimates aid the breeder in choosing the most efficient selection method and breeding System. With body weight being an important trait to all breeders, it seems desirable to consolidate all heritability estimates of this trait, as well as to report those that have been calculated at the Oklahoma Agricultural Experiment Station during the past four years. REVIEW OF THE LITERATURE Heritability estimates of body weight, as found in the literature, are summarized by age and method of estimation in Table 1. A simple, arithmetic average of these estimates by ages show that the heritability of body weight at broiler . . .
Article
1. The aim of present study was to estimate genetic parameters for average egg weight (EW) and egg number (EN) at different ages in Japanese quail using multi-trait random regression (MTRR) models. 2. A total of 8534 records from 900 quail, hatched between 2014 and 2015, were used in the study. Average weekly egg weights and egg numbers were measured from second until sixth week of egg production. 3. Nine random regression models were compared to identify the best order of the Legendre polynomials (LP). The most optimal model was identified by the Bayesian Information Criterion. A model with second order of LP for fixed effects, second order of LP for additive genetic effects and third order of LP for permanent environmental effects (MTRR23) was found to be the best. 4. According to the MTRR23 model, direct heritability for EW increased from 0.26 in the second week to 0.53 in the sixth week of egg production, whereas the ratio of permanent environment to phenotypic variance decreased from 0.48 to 0.1. Direct heritability for EN was low, whereas the ratio of permanent environment to phenotypic variance decreased from 0.57 to 0.15 during the production period. 5. For each trait, estimated genetic correlations among weeks of egg production were high (from 0.85 to 0.98). Genetic correlations between EW and EN were low and negative for the first two weeks, but they were low and positive for the rest of the egg production period. 6. In conclusion, random regression models can be used effectively for analysing egg production traits in Japanese quail. Response to selection for increased egg weight would be higher at older ages because of its higher heritability and such a breeding program would have no negative genetic impact on egg production.
Article
Data from 54 selection rounds (SRs) for growth rate in a selection line from Cobb breeding company was used for analysis with the objective to investigate the extent of sex by genotype interaction for body weight (BW) in a commercial broiler chicken population, estimate genetic parameters of BW at three different ages, and study the development in genetic variance for BW over several generations of selection. BW was measured at three different ages (t, t-4 and t-7 days) in different birds of both sexes. For the first 39 SRs, BW was recorded at t days, however, as selection continued the birds reached desired BW earlier and the weighing age was changed to t-4 days for 7 SRs and then to t-7 days for the last 8 SRs. A bivariate random regression model regarding BW in each sex as different traits using segmented linear splines and heterogeneous residual variance was used to estimate genetic parameters of BW across the selection trajectory via Gibbs sampling. The genetic variance and the heritability estimates of BW in males and females were found to be different at all three ages. The genetic correlation between BW measured in males and females was less than unity and decreased further as age at weighing was increased. This illustrated that BW in the two sexes should be considered as different but correlated traits and models for BW should account for a decreasing genetic correlation as the distance between the weighing ages increased. An increased genetic variance of BW along the selection trajectory in the first period was observed and genetic correlations between BW measured in different SRs gradually decreased as the distance between SRs increased. Genetic correlations lower than unity between BW at early and late SRs in the first period indicated that the genetic background of BW has gradually changed along the selection trajectory. This change probably contributes significantly to maintaining large levels of genetic variance in highly selected broiler populations.