1. Development of Genomic GMACE
Pete Sullivan, CDN
&
Paul VanRaden*, USDA

2. Introduction Genomics is here now
National EBVs are being replaced by GEBVs
GEBVs may not be strictly national
Countries may share data, or use MACE
Genetic tests may be repeated (DGAT), shared or sold among countries

3. Introduction MACE assumption: Genomics:
National data sets are independent, i.e. separate recorded cow populations
Sire with EBV from 100 daughters in 3 countries has 300 daughters in total
Genomics:
Sire can have GEBV from 100 daughters in 3 countries with only 100 daughters in total
MACE must evolve to GMACE

4. Objectives Overall objectives: Today:
Update MACE model for genomic input data from participating countries
Create GMACE software for application as early as 2010
Today:
Introduce GMACE with some examples
Preliminary results from software tests
Reactions and feedback from members

5. Methods
De-regression
MACE

6. Methods
De-regression
MACE
GMACE

7. Methods D is a diagonal matrix
Residual variances of de-regressed proofs
E = D plus genomic covariances from shared data
% common (shared) data
Max correlation between genomic predictions
( = %EDC from genomics)
Genomic portion of variance

8. Single Sire Example GMACE for a single sire
GEBV in 3 countries (A, B, C)
EDC from genomics = 20
No data in 4th country D
All rg = .90, h2 = .33,
as a young bull (EDC=0+20, ɣ=100%)
with 1st crop proofs in A, B and C (EDC=100+20, ɣ=17%)

9. Single Sire Example - Daughter equivalents
Young Bull (ɣ=100%)
1st Crop Proofs (ɣ=17%)
Countries
A,B,C D
National 20
120
MACE 42 30
172 63
GMACE
c=0.0
c=0.5 25 19
163 61
c=1.0 14
155 60

10. Single Sire Example - GEBV ( )
Young Bull (ɣ=100%)
1st Crop Proofs (ɣ=17%)
Countries
A,B,C D
National -
3.0 – 3.5
MACE
4.3 – 4.4
4.2
3.3 – 3.6
3.3
GMACE
c=0.0
c=0.5
3.6 – 3.8
3.6
c=1.0
3.1

11. Brown Swiss Data - 9 countries
Used Pedigree and EBV from April 2009 Interbull files to simulate bull genotypes
50,000 SNP and 10,000 QTL
VanRaden (2009)
8073 proven bulls
120 young bulls sampled in U.S.A. used to test genomic predictions

12. Brown Swiss Data Young USA Bull Reliability
GMACE

13. Brown Swiss Data Young USA Bull Reliability
GMACE reliability very close to global genomic evaluation model
If “c” is correct
No shared data for GEBV in this study (i.e. c=0.0)
Some error in “c” seems O.K. (c=0.5 still good)
How robust is GMACE?
c=? %EDC from genomics ( =?)

14. GMACE – next steps
Need to test GMACE when data are shared among countries for national GEBVs (c>0)
Need to extend MACE / MT-MACE reliability approximation for GMACE
Needs multi-country progeny absorption if c>0
May need MT-GMACE (e.g. for fertility traits)
Are national cow EBVs useful for GMACE?
Question also for MACE, but
More important for young genotyped bulls in GMACE

15. Conclusions
Software to extend from national to international GEBVs nearing completion
Application by Interbull in 2010? 2011?
A few developments and validation tests are still needed for the model and software
Enhancements may be added after the initial application
Extension to MT-GMACE
Model refinements

16. Acknowledgements Interbull genomics task force Georgios Banos
Mario Calus
Vincent Ducrocq
João Dϋrr
Hossein Jorjani
Esa Mäntysaari
Zengting Liu

17. Thank You!