implementation of genomic selection in dairy cattle
TRANSCRIPT
![Page 1: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/1.jpg)
21112005
Implementation of genomic selection in dairy cattle
breeding schemes
Marie Lillehammer (Nofima)
Anna Sonesson (Nofima)
Theo Meuwissen (Norw.Univ.Life Sci.)
![Page 2: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/2.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Introduction
GEBV are available in many countries
– GBLUP or blended GBLUP-TBLUP
– BayesA/B/C
Expectations high: accurate EBV of:• Young genotyped animals
• Nonrecorded animals (trait nor pedigree)
– Difficult /costly traits
• Animals living in a different environment
• Low heritability traits
• More sustainable breeding scheme
Institute for Animal and Aquacultural Sciences
2
![Page 3: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/3.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
AIM:
Compare alternative designs for implementing GEBV in Norwegian Red breeding schemes
Use stochastic simulation where accuracy of GWEBV is a result of the design
– Deterministic: accuracy is input/ independent of design (Schaeffer et al., 2007; Kønig et al., 2009)
Institute for Animal and Aquacultural Sciences
3
![Page 4: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/4.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Methods: simulation of base population
2,000 generations
Ne=200 (Fisher-Wright idealised pop.)
30 BTAs of 1 Morgan each (106 bp)
Mutation 10-8/bp (infinite sites mutation mod)
Recombination 10-8/bp
3,000 random SNPs with MAF>.05 => QTL
– QTL effects from reflected exponential distrib.
15,000 SNPs with highest MAF =>marker
– Marker ≠ QTL
Institute for Animal and Aquacultural Sciences
4
![Page 5: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/5.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Simulation of breeding scheme
– Not possible to simulate entire Norwegian Red pop.
– Reduced size of simulated population• Number of selected males the same (in SD and SS)
• Selected fractions identical (and selection intensities)
• Selection steps for other traits: omitted
• Non-GS larger scheme: similar G and F
– Test-daughters were not (individually) simulated• Reduces population size
Institute for Animal and Aquacultural Sciences
5
![Page 6: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/6.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Preselection of young bulls (PS)
Institute for Animal and Aquacultural Sciences
6
750 male calves 750 female calves
125 young bulls
6700 female calves
12 elite sires1500 elite dams
6700 random females
Population of females
PT
GS/PedIndx
90%5-10%
~20,000
Selection (TBLUP, unless stated otherwise)
•60/80/100
•More progeny/bull (A)
•Fewer test-dghtrs (B)
![Page 7: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/7.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Full genomic selection scheme (GS)
Institute for Animal and Aquacultural Sciences
7
750 male calves 750 female calves
125 young bulls
6700 female calves
12 elite sires1500 elite dams
6700 random females
Population of females
PT
GS/PedIndx
90%5-10%
~20,000
Selection (TBLUP, unless stated otherwise)
GS
• 20 / 30 / 40
![Page 8: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/8.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Traits & GEBV
Institute for Animal and Aquacultural Sciences
8
∑ =+= 3000
1 2211j jijjiji gxgxTBV
•Yi=TBVi+ei
•ei~N(0,Ve)
•Ve is adjusted so that h2 is .01, .15 or .30
•Trait recording: at 2-yr-old females (gen.interv. =
3yr)
•Progeny test: 5-yr-old sires (gen.interv. = 6yr)
•GS: only applied to young-bulls;
•GBLUP (BLUP of marker effects; no blending): ∑ =++= n
j ijiji eaXy1
μ ∑ == n
j jiji axGEBV1
![Page 9: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/9.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Simulation parameters:
100 replicates
Start population:
– no genetic trend
– Training set of 3000 progeny tested bulls
Results for yr 13-19
– Accuracy : yr 19
Institute for Animal and Aquacultural Sciences
9
![Page 10: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/10.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Results: preselection schemes
Institute for Animal and Aquacultural Sciences
10
ΔF ΔG AccConv. 1 1 xxPS_125 1.13 1.29 0.7PS_60A 1.15 1.29 0.68PS_60B 1.06 1.27 0.66
![Page 11: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/11.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Full GS schemes
Institute for Animal and Aquacultural Sciences
11
ΔF ΔG AccConv. 1 1 xxGS_12 1.65 1.52 0.61GS_30 0.8 1.42 0.63GS_40 0.61 1.37 0.63
![Page 12: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/12.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Effect of h2 on G
Institute for Animal and Aquacultural Sciences
12
h2 0.01 0.05 0.3Conv. 1 1 xxPS_125 1.32 1.31 1.29PS_60A 1.39 1.3 1.27GS_12 1.87 1.59 1.5GS_40 1.59 1.42 1.36
![Page 13: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/13.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Trend in accuracy
Institute for Animal and Aquacultural Sciences
13
Accuracy GEBV - h2=0.15
0.6
0.62
0.64
0.66
0.68
0.7
0.72
10 11 12 13 14 15 16 17 18 19 20
Year
PS_125 PS_60A PS_60B GS_12 GS_20 GS_40
![Page 14: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/14.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Accuracy of GEBV
Institute for Animal and Aquacultural Sciences
14
Accuracy GEBV - h2=0.15
0.6
0.62
0.64
0.66
0.68
0.7
0.72
0 20 40 60 80 100 120 140
N umb er o f g eno t yp ed sires wit h p ro g eny
GS PS_A PS_B
•GS less accur. at same no. of training bulls
•In GS training bulls are from generation t-2
![Page 15: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/15.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
GS versus PS
GS highest G (+18%)
GS highest F (+46%)
– Short generation interval
– Alleviated by selecting 30 elite sires: F : -20% and G =+10%
GS is cheapest (no progeny testing)
BUT:
– GS is furthest from conventional scheme (risk)
– Has lowest accuracy (farmer acceptance)
Institute for Animal and Aquacultural Sciences
15
![Page 16: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/16.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Low heritability traits
Schemes rank the same
GS gives relative more extra G
More sustainable breeding scheme
– Functional traits more easily improved by GS
– BUT: still need large scale recording
Institute for Animal and Aquacultural Sciences
16
![Page 17: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/17.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Including females
For selection:
– Reduce generation interval in DS path
– Avoids preferential treatment problem
For training:
– Need huge numbers • h2=.15 => 6 times as many
Work in progress (Noirin McHugh)
– Irish Dairy cattle
Institute for Animal and Aquacultural Sciences
17
![Page 18: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/18.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Other traits : not large scale recorded
Idea for difficult trait:
– Perform experiment (~2000 records)
– Estimate SNP effects
– Select for SNP effects for ever after
– Accuracy reduces markedly over time• (Muir, 2007; Habier etal, 2007; Sonesson et al., 2009)
• Need continuous scheme for updating SNP effects
• Not when using genome sequence (Meuwissen&Goddard,2010)
Institute for Animal and Aquacultural Sciences
18
![Page 19: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/19.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Conclusions:
Full GS scheme is best– Highest G and F
– Can be tuned to have low F
– Cheapest
– BUT: most risky
Pre-selection scheme:
– Close to current scheme
– Can safe some costs by reducing progeny test• Beneficial for farmer
– High accuracy
Institute for Animal and Aquacultural Sciences
19
![Page 20: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/20.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Conclusions (2)
Increased F
– shorted generation interval in GS scheme
– also in PS scheme
Accuracy remained high when h2 decreased
– Unless h2 became very low
– Lower in GS than PS• One generation more between training and candidates
• Kept on decreasing in scheme with few elite sires
Institute for Animal and Aquacultural Sciences
20
![Page 21: Implementation of genomic selection in dairy cattle](https://reader031.vdocuments.mx/reader031/viewer/2022012519/6193994a3b144663f043f77e/html5/thumbnails/21.jpg)
NO
RW
EGIAN
UN
IVERSITY OF LIFE SCIEN
CES
www.umb.no
Conclusions (3)
More sustainable breeding schemes
– Low h2 traits easier to improve
– However in practice r2 of low h2 trait disappointing (Luan et al., 2010)
– Still need large scale recording• Re-train SNP effects
Institute for Animal and Aquacultural Sciences
21