Transcript Estimation of genetic parameters between single

Estimation of genetic parameters
between
single-record and multiple-record traits
using ASREML
Sansak Nakavisut
Principle supervisor: Dr Ron Crump
Co-supervisor: Dr Hans Graser
Outline
• Introduction
• Univariate analysis of single-record traits (eg ADG)
• Bivariate analysis of 2 (single-record) traits (eg ADG & FI)
• Univariate analysis of multi-record traits (eg NBA)
• Bivariate analysis of single & multi-record traits
• Problems / solutions
• Demonstration
Introduction
• Aim: to demonstrate how to estimate genetic
parameters from a more complex model using
ASREML
• Data set up to match the model
• Concept of multi-record traits (repeated
measurements of the same traits)
• Problems you may encounter in bivariate analysis
• and solutions
Univariate analysis of single-record trait
• One record per animal during the life time
• eg BW, TN, BF, FCR, ADG, FI, etc.,
• Genetic parameters; ha2, hm2 , c2, ram ,
estimable and reliable or not, depending on
data structure
• Std. error will reflect data structure (Qnt Qly)
Example: univar. Of ADG
“ADG.as” (COMMAND FILE)
Analysis of production traits
Anim !P
Sire !P
Dam !P
Br 2 !A
Sex 2 !A
HYS 2 !A
ADG
FI
test.ped !ALPHA
demo.dat !MVINCLUDE !DOPART $1
!PART 1
ADG ~ mu Br Sex !r Anim !f HYS
“demo.dat” (DATA FILE)
TT2H0453
TT2H0456
TT2H0483
TT2H0484
TK1H0246
TK1H0239
TK1H0228
TT2H0495
TT1H7156
TT1H7157
TK1H0226
TK1H0224
…
TT2E4326
TT2E4326
TT2C1953
TT2C1953
IR1E0003
IR1E0030
TK1G0056
TT2E4326
TT1F4957
TT1F4957
TK1G0056
TK1G0056
TT2E2627
TT2E2627
TT2E4315
TT2E4315
IR1E0137
IR1E0139
TK1G0042
TT2F4908
TT1F4990
TT1F4990
TK1G0042
TK1G0042
LR
LR
LR
LR
LW
LW
LW
LR
LW
LW
LW
LW
“test.ped” (PEDIGREE FILE)
RC3K2178
RC3L1647
RC3L2652
RC3L2731
RC3L3109
RC3M3292
RC3M3424
……
. .
. .
. .
RC3-0017
RC3-1244
RC3K2210
RC3L2765
RC3-278-9
RC3-1274
RC3-1798
RC3-0567
M
F
F
F
M
M
M
F
M
F
M
M
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK943
TK943
497.75
515.17
490.48
417.43
561.45
538.46
476.68
456.94
528.25
494.59
505.18
466.32
1.63
1.88
1.85
1.77
1.76
1.71
1.56
1.82
1.93
1.69
1.79
1.61
Results: univariate analysis (ADG)
“ADG1.asr”
Source
Anim
Variance
Model
9621
7665
terms
9621
7530
Gamma
1.10486
1.00000
Component
1129.25
1022.08
Comp/SE
16.25
23.70
“ADG1.pin”
F Vp 1 + 2
H h2 1 3
“ADG1.pvc”
3 Vp
h2
1
2151.
= Anim
44.39
1/Vp 1
3=
0.5249
0.0245
% C
0 P
0 P
Bivariate analysis of single-record traits
• Joint analysis of 2 traits (eg ADG and FI)
• Genetic parameter estimates;
• h2 of trait 1(ADG) (accounted for trait 2(FI))
• h2 of trait 2(FI) (accounted for trait 1(ADG))
• PLUS
• rg12, re12, rp12
Bivariate analysis of ADG and FI
“ADG.as” (COMMAND FILE)
Analysis of production traits
Anim !P
Sire !P
Dam !P
Br 2 !A
Sex 2 !A
HYS 2 !A
ADG
FI
test.ped !ALPHA
demo.dat !MVINCLUDE !DOPART $1
!PART 2
ADG FI ~ Trait Tr.Br Tr.Sex !r
Tr.Anim !f Tr.HYS
1 2 1
0
Tr 0 US 1 0.1 1 !GP
Tr.Anim 2
Tr 0 US 1 0.1 1 !GP
Anim
R
ADG
ADG
1
FI
0.1
1
G
ADG
FI
ADG
1
FI
0.1
FI
1
asreml –rs4 ADG 2
Results: bivariate analysis
1
2
3
4
5
6
“ADG2.asr”
Source
Model terms
Gamma
Component
Comp/SE
Residual
UnStruct
1
1022.23
1022.23
23.71
Residual
UnStruct
1
2.25786
2.25786
14.78
Residual
UnStruct
2 0.259174E-01 0.259174E-01 29.36
Tr.Anim
UnStruct
1
1128.76
1128.76
16.25
Tr.Anim
UnStruct
1
2.28471
2.28471
9.78
Tr.Anim
UnStruct
2 0.165168E-01 0.165168E-01 13.14
Covariance/Variance/Correlation Matrix UnStructured
1022.
0.4387
2.258
0.2592E-01
Covariance/Variance/Correlation Matrix UnStructured
1129.
0.5291
2.285
0.1652E-01
“ADG2.pin”
F
F
F
H
H
R
R
VpADG 1+4
VpFI 3+6
Covp12 2+5
h2_ADG 4 7
h2_FI 6 8
rg 4 5 6
rp 7 9 8
% C
0 P
0 P
0 P
0 P
0 P
0 P
“ADG2.pvc”
7 VpADG 1
8 VpFI 3
9 Covp12 2
h2_ADG
h2_FI
rg
rp
2151.
44.38
0.4243E-01 0.8272E-03
4.543
0.1506
= Tr.Anim
4/VpADG 1
7=
= Tr.Anim
6/VpFI 3
8=
= Tr.Anim /SQR[Tr.Anim *Tr.Anim ]=
= Covp12 /SQR[VpADG 1*VpFI 3 ]=
0.5248
0.3892
0.5291
0.4755
0.0245
0.0248
0.0361
0.0109
Univariate analysis of multi-record trait
• When rg between 2 measurements is 1 or close to
unity, “repeated measurements of the same trait”,
within-individual variance is caused by temporary
differences of environment
• Otherwise, we should treat them as “2 different
traits” they are not under the same genetic control
• Genetic parameter estimates;
• ha2, hm2 , c2, ram if data structure allows
• PLUS repeatability (r)
data set-up for multi-record trait (NBA)
For repeatability model
An S D Fixed
A
A
A
A
B
B
C
C
C
.
.
.
1
1
1
1
3
3
3
3
3
2
2
2
2
4
4
5
5
5
.
.
.
.
.
.
.
.
.
Par
1
2
3
4
1
2
1
2
3
Treat NBA1-4 as different traits
NBA
10
11
10
13
9
12
10
11
13
An S D Fixed NBA1 NBA2 NBA3 NBA4
A
B
C
.
.
.
1 2
3 4
3 5
.
.
.
10
9
10
11
12
11
10
.
13
13
.
.
Example: repeatability model (NBA)
“NBA.as”
“NBA.dat”
Analysis of NBA (repeatibility model)
CA1G0037 CA1F33910 CA1F0040 LW F TK951 11
Anim !P
CA1G0037 CA1F33910 CA1F0040 LW F TK952 12
CA1G0037 CA1F33910 CA1F0040 LW F TK953
5
Sire !P
CA1G0038 CA1F33910 CA1F0040 LW F TK943
9
Dam !P
CA1G0038 CA1F33910 CA1F0040 LW F TK953
5
CA1G0038 CA1F33910 CA1F0040 LW F TK961
6
Br 2 !A
CA1G0038 CA1F33910 CA1F0040 LW F TK963
3
Sex 2 !A
CA1G0038 CA1F33910 CA1F0040 LW F TK972
2
HYS 2 !A
NBA
CA1G0038 CA1F33910 CA1F0040 LW F TK973 12
CA1G0050 CA1E0226
.
.
repro.ped !ALPHA
NBA.dat !REPEAT !MAXIT 50 !MVINCLUDE
NBA ~ mu Br !r Anim ide(Anim) !f HYS
.
CA1E0040 LW F TK952 10
Results: repeatability model
“NBA.asr”
Source
Anim
ide(Anim)
Variance
Model
5421
5421
10948
terms
5421
5421
10690
Gamma
Component
0.112836
0.644805
0.652040E-01 0.372611
1.00000
5.71454
Comp/SE
6.04
3.69
62.25
%
0
0
0
C
P
P
P
“NBA.pin”
F
F
H
H
Vp 1+2+3
repeat 1+2
h2 1 4
r 5 4
“NBA.pvc”
h2 = Anim
1/Vp
r = repeat 5/Vp
1 4= 0.0958
1 4= 0.1511
0.0155
0.0108
Bivariate analysis of single & multi-record traits
•Data Set-up
•Command file( .as file) & Model
•Which terms to estimate what
•Problems / Solutions
•Demonstration
Data Set-up: single-multi records
Anim Br Sex
.
.
.
YL2O3774 LR F
YL2O3774 LR F
YL2O3774 LR F
YL2O3779 LR F
YL2O3779 LR F
YL2O3798 LR F
YL2O3798 LR F
YL2O3800 LR F
TT2H0453 LR M
TT2H0456 LR F
TT2H0483 LR F
TT2H0484 LR F
TK1H0246 LW M
TK1H0239 LW M
TK1H0228 LW M
TT2H0495 LR F
TT1H7156 LW M
TT1H7157 LW F
.
.
.
HYS NBA_ADG NBA_FI Tr
YL022
YL031
YL032
YL022
YL031
YL023
YL031
YL023
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
TK942
9
9
12
12
11
11
10
10
8
8
11
11
10
10
7
7
497.75 1.63
515.17 1.88
490.48 1.85
417.43 1.77
561.45 1.76
538.46 1.71
476.68 1.56
456.94 1.82
528.25 1.93
494.59 1.69
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
Multi-records of NBA
Single-record of
ADG & FI
Command file: single-multi records
Analysis of NBA and ADG traits
Anim !P
Br 2 !A
Sex 2 !A
Ve1 = residual V. left from
HYS 2 !A
NBA_ADG
Ve2 - Ve1 - Cov e12
NBA_FI
Tr 2
demo.ped !ALPHA
demo1.dat !REPEAT !MAXIT 50 !MVINCLUDE
the model
Cov e12
NBA_ADG ~ Tr Tr.Br at(Tr,2).Sex !r Tr.Anim ide(Anim) !GU,
at(Tr,1).ide(Anim) uni(Tr,2) !GU !f Tr.HYS
0 0 1
Tr.Anim 2
Tr 0 US 1 0.1 1 !GP
G structure
Anim
Concept: single-multi records
R
NBA1
NBA1
Ve1
NBA2
NBA3
NBAn
NBA2
Cov11
Ve1
NBA3
Cov11
Cov11
Ve1
NBAn
Cov11
Cov11
Cov11
Ve1
ADG
Cov12
Cov12
Cov12
Cov12
ADG
Ve2
R
NBA
NBA
Ve1
ADG
Cov12
ADG
Ve2
Results: single-multi records (NBA-ADG)
Source
Model
terms
Gamma
Component
Comp/SE
% C
1 ide(Anim)
11414
11414
0.764115
4.36474
2.25
0 U
2 at(Tr,1).ide(Anim)
11414
11414 -0.701052
-4.00452
-2.06
0 U
3 uni(Tr,2)
18613
18613
176.829
1010.08
23.44
0 U
4 Variance
18613
18220
1.00000
5.71215
62.25
0 P
5 Tr.Anim
UnStruct
1
0.116258
0.664080
6.21
0 P
6 Tr.Anim
UnStruct
1 -0.856536
-4.89266
-2.27
0 P
7 Tr.Anim
UnStruct
2
1133.63
16.29
0 P
198.460
“demo.pin”
F
F
F
F
F
H
H
H
R
R
(8)Ve2 1+3+4
(9)Vp1 5+1+2+4
(10)Vp2 7+8
(11)Covp12 1+6
(12)repeat1 1+2+5
h12 5 9
r1 12 9
h22 7 10
rg 5 6 7
rp 9 11 10
“demo.pvc”
h12
r1
h22
rg
rp
=
=
=
=
=
Tr.Anim
5/(9)Vp1
9=
0.0986
(12)repe 12/(9)Vp1
9=
0.1521
Tr.Anim
7/(10)Vp2
10=
0.5263
Tr.Anim /SQR[Tr.Anim*Tr.Anim]= -0.1783
(11)Covp/SQR[(9)Vp1 *(10)Vp2]= -0.0044
0.0154
0.0108
0.0245
0.0777
0.0173
Comparison: univariate – (semi)bivariate
“univariate analysis of NBA”
Source
Anim
ide(Anim)
Variance
Model
5421
5421
10948
terms
5421
5421
10690
Gamma
Component
0.112836
0.644805
0.652040E-01 0.372611
1.00000
5.71454
Comp/SE
6.04
3.69
62.25
%
0
0
0
C
P
P
P
“(semi)bivariate analysis of NBA-ADG”
Source
Model
terms
Gamma
Component
Comp/SE
% C
ide(Anim)
11414
11414
0.764115
4.36474
2.25
0 U
at(Tr,1).ide(Anim)
11414
11414 -0.701052
-4.00452
-2.06
0 U
uni(Tr,2)
18613
18613
176.829
1010.08
23.44
0 U
Variance
18613
18220
1.00000
5.71215
62.25
0 P
Tr.Anim
UnStruct
1
0.116258
0.664080
6.21
0 P
Tr.Anim
UnStruct
1 -0.856536
-4.89266
-2.27
0 P
Tr.Anim
UnStruct
2
198.460
1133.63
16.29
0 P
terms
9621
7530
Gamma
1.10486
1.00000
Component
1129.25
1022.08
Comp/SE
16.25
23.70
% C
0 P
0 P
“univariate analysis of ADG”
Source
Anim
Variance
Model
9621
7665
Estimates: univariate – (semi)bivariate
“NBA.pvc”
h2 =
r =
Anim
1/Vp
repeat 5/Vp
1 4 =
1 4 =
0.0958
0.1511
0.0155
0.0108
“demo.pvc”
h12 = Tr.Anim
r1 = (12)repe
5/(9)Vp1
12/(9)Vp1
9=
9=
0.0986
0.1521
0.0154
0.0108
10=
0.5263
0.0245
= Tr.Anim /SQR[Tr.Anim*Tr.Anim]= -0.1783
= (11)Covp/SQR[(9)Vp1 *(10)Vp2]= -0.0044
0.0777
0.0173
h22 = Tr.Anim
rg
rp
7/(10)Vp2
“ADG1.pvc”
h2
= Anim
1/Vp
1
3=
0.5249
0.0245
Problems / solutions
• Convergence failed
• Rescaling variable(s) to have similar Vp for both traits
• eg FI = 1.1 kg/d => 11 (/10)kg/d to match NBA of 10 pigs/litter
• Re-run 99% solved
• Rescaling may change missing values to zero
• eg “.” x10 = 0
• Be careful, keep missing value as it is
• Bizarre Outputs from similar analyses
• Check the order of terms in .asr carefully
• Even with the same set-up of files, ASREML may report terms in
different orders than some previous runs
• Re-order components in .pin file to match .asr file
Example: problem of order
Source
at(Tr,2).Litter
ide(Anim)
at(Tr,1).ide(Anim)
uni(Tr,2)
Variance
Tr.Anim
Tr.Anim
Tr.Anim
F
F
F
F
F
F
F
H
H
H
H
R
R
R
E11 5
E22 2+4+5
E12 2
P11 2+3+5+6
P22 8+1+10
P12 2+7
S11 2+3+6
c2 1 13
r11 15 12
h21 6 12
h22 8 13
rg 6 7 8
rp 12 14 13
re 9 11 10
Component
276.084
3.34576
-3.11344
417.221
6.16018
0.785246
-4.33941
174.395
Source
at(Tr,2).Litter
uni(Tr,2)
ide(Anim)
at(Tr,1).ide(Anim)
Variance
Tr.Anim
Tr.Anim
Tr.Anim
F
F
F
F
F
F
F
H
H
H
H
R
R
R
E11 5
E22 2+3+5
E12 3
P11 3+4+5+6
P22 8+1+10
P12 3+7
S11 3+4+6
c2 1 13
r11 15 12
h21 6 12
h22 8 13
rg 6 7 8
rp 12 14 13
re 9 11 10
Component
35.6195
55.0355
0.912272
-0.669691
6.16330
0.772876
-1.41296
28.7313
Demonstration