AFEchidna可以运行复杂模型,简单示例如下:
pm3<-echidna(fixed = cbind(weanwt,weight)~Trait:(year+sex+weanage+pen),
random=~str(~Trait:(nrm(pig)+nrm(dam)),~us(4):nrm(pig)),
residual=~idv(units):us(Trait),
mulT = TRUE,
es0.file='pig_data.es0')
方差分量结果如下:
> Var(pm3)
Term Sigma SE Z.ratio
1 Residual 1.00000 0.00000 Inf
2 us(4):nrm(pig);us(4) 0.42680 0.31681 1.34717970
3 us(4):nrm(pig);us(4) 0.66695 2.90190 0.22983218
4 us(4):nrm(pig);us(4) 184.65000 52.62300 3.50892195
5 us(4):nrm(pig);us(4) -1.04540 0.27391 -3.81658209
6 us(4):nrm(pig);us(4) -5.75300 3.68040 -1.56314531
7 us(4):nrm(pig);us(4) 4.40180 0.39105 11.25636108
8 us(4):nrm(pig);us(4) -2.08060 2.17520 -0.95650975
9 us(4):nrm(pig);us(4) 1.64770 33.44500 0.04926596
10 us(4):nrm(pig);us(4) 5.12130 2.54910 2.00906202
11 us(4):nrm(pig);us(4) 164.52000 30.91500 5.32168850
12 id(units):us(Trait);us(Trait) 2.08390 0.17053 12.22013722
13 id(units):us(Trait);us(Trait) 10.31300 1.60630 6.42034489
14 id(units):us(Trait);us(Trait) 384.59000 29.39300 13.08440785
上述的结果解读,比较费劲。可结合Echidna的原始结果:
cat(pm3$components)
带下标的结果如下:
> cat(pm3$components)
Model_Term Order Gamma Sigma Z_ratio %C
Trait.nrm(pig) 5414 effects
us(4)_V 1 1 4 0.426800 0.426800 1.35 0 P
us(4)_C 2 1 4 0.666954 0.666954 0.23 0 P
us(4)_V 2 2 4 184.655 184.655 3.51 0 P
us(4)_C 3 1 4 -1.04544 -1.04544 -3.82 0 P
us(4)_C 3 2 4 -5.75303 -5.75303 -1.56 0 P
us(4)_V 3 3 4 4.40178 4.40178 11.26 0 P
us(4)_C 4 1 4 -2.08062 -2.08062 -0.96 0 P
us(4)_C 4 2 4 1.64771 1.64771 0.05 51 P
us(4)_C 4 3 4 5.12134 5.12134 2.01 0 P
us(4)_V 4 4 4 164.520 164.520 5.32 0 P
id(units).us(Trait) 5112 effects
us(Trait)_V 2 2.08394 2.08394 12.22 0 P
us(Trait)_C 2 10.3134 10.3134 6.42 0 P
us(Trait)_V 2 384.593 384.593 13.08 0 P
上述结果中的id(units):us(Trait);us(Trait)
项容易理解,为性状1和性状2的us()方差结构。
而us(4):nrm(pig);us(4)
比较费解。根据模型中random项=str(~Trait:(nrm(pig)+nrm(dam)),~us(4):nrm(pig))
,其中us(4)表示该方差结构为4个方差项的us结构,前两项为Trait:nrm(pig),后两项为Trait:nrm(dam),由此推出项1为Trait1:nrm(pig),项2为Trait2:nrm(pig),项3为Trait1:dam(pig),项4为Trait2:dam(pig)。项数对应order下标,据此,就可获悉各项方差分量的含义。
参考文献:
Zhang WH, Wei RY, Liu Y, Lin YZ. AFEchidna is a R package for genetic evaluation of plant and animal breeding datasets. BioRxiv. DOI: 10.1101/2021.06.24.449740.