Aug 30, 2013 - Progeny tested bucks each year ... AI bucks. Alpine or Saanen. CD asc. : 38%. 2011: Illumina goat SNP50 Bead .... 0.2 Charolais-Angus (600kb)
First Steps towards Genomic Selection in French Dairy Goat Céline Carillier, Christèle Robert-Granié, Hélène Larroque INRA-SAGA Toulouse 64th EAAP session, 26th - 30th August, 2013, Nantes, France
Few males progeny tested each year in French dairy goats Progeny tested bucks each year 40 Alpine
30 Saanen 100 daughters /year with performances
25 Alpine
CDasc : 38%
AI bucks
15 Saanen
> 5.5 years
sire-son pathway
Alpine or Saanen
2011: Illumina goat SNP50 Bead Chip .02
Not ideal population structure: prediction reliability similar to dairy sheep linkage disequilibrium
Genomic evaluation
interest of adding females in reference population
.03
Small multi-breed population genotyped with females 46 959 SNP after quality control 1993
677 males 384 Alpines
293 Saanens
1,985 females
QTL detection
2008 2009
1,243 Alpines
742 Saanens
2010
148 males not progeny tested yet 87 Alpines
2011 61 Saanens .04
Medium linkage disequilibrium in dairy goats, lower in multi-breed population than in each breed 0,2
Alpine+Saanen Alpine Saanen
Average r²
0,17 0,15 0,14
Other species: 0.18 - 0.3 in dairy cattle 0.11 - 0.17 in French dairy sheep
0,1
0,05
0
0
60
500
1000
1500
2000
Marker distance (kb) .05
Alpine and Saanen : two different breeds according to LD persistence between both 1 Manech red faced - Manech black faced Manech red faced - Lacaune Alpine - Saanen
0,9
Correlation of r
0,8 0,7
0,6 0,58 0,5 0,4 0,3 0,2
0,08
0,1
0 0 70
500 600
1000 Marker distance (kb)
1500
2000
.06
What are the model and the data used in this study? phenotypes and pedigree official genetic evaluation
genotypes Performances corrected for fixed effects YD, DYD pedigree
genomic BLUP
y X Zu e
β : breed effect u : breeding values
- breeding values - prediction error variance (PEV) .07
How we calculate prediction reliability? reference population training set (425 males born between 1993 and 2005) DYD in 2008
genotypes
predicting equation
r² ( DYDobs, GEBV)= prediction reliability
validation set (252 males born between 2006 and 2009)
genotypes
DYD in 2013
genomic breeding values
.08
Prediction reliability in validation population close to the one in Normande dairy cattle breed Correlations DYD*GEBV (%)
60 50
Other French results: - Holstein cow (milk): 60% - Normande cow (milk) : 36% - Lacaune sheep (milk): 45%
40 30 20 10 0
.09
No males with daughters born after females genotyped 46 959 SNP after quality control 1993
677 males 384 Alpines
293 Saanens
1,985 females
QTL detection
2008 1,243 Alpines
742 Saanens
2010 2011
148 males not progeny tested yet 87 Alpines
2009
61 Saanens .010
How we calculate accuracy of young buck breeding values? reference population 677 males DYD in 2013
genotypes
+
females genotypes YD in 2013
predicting equations
candidates (148 young males)
genotypes
u 2 PEV accuracy u 2
breeding values prediction error
breeding values prediction error
accuracies without females
accuracies with females
𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦𝑤𝑖𝑡ℎ 𝑓𝑒𝑚𝑎𝑙𝑒𝑠 − 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑓𝑒𝑚𝑎𝑙𝑒𝑠 𝑔𝑎𝑖𝑛 = 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑓𝑒𝑚𝑎𝑙𝑒𝑠
.011
Gain in accuracy (%)
Improvement of accuracies when adding 1,985 females in reference population of 677 males 10
8 6 4 2 0
.012
Genomic selection is difficult to apply due to small population size, two breeds and population structure population 2 breeds, small size
all males
half-sibs females
good model fit improvement of GEBV accuracy with females
What is the most accurate model adapted to multiple breeds?
questions?
What are the females to be genotyped? .013
Appendix
.014
Difficulties to consider Alpine and Saanen as a single breed common ancestor
Saanen
Axis 1-2
Alpine
correlation between allele frequencies : Males
0.6852
Females
0.8455
Both
0.8605
ACP on genotypes, factor: breed
.015
Young buck breeding value accuracies not as high as expected Reference population: 677 males and 1,985 females
breeding values accuracies
0,35 0,25 0,15 0,05
-0,05
.016
How we calculate prediction reliabilty? reference population training set (425 males born between 1993 and 2006) DYD in 2008
genotypes
predicting equation
r² ( DYDobs, EBV) GEBV)= = prediction reliability
validation set (252 males born between 2007 and 2009)
genotypes
DYD in 2013
Gain in reliability
breeding values
r²(DYDobs, GEBV) - r²(DYDobs, EBV) r²(DYDobs, EBV)
.017
Correlations (%) between DYD and GEBV
Gain in reliability with genomic information lower than in other species for milk production traits 25
Other species: -Lacaune sheep (milk): 15% (udder): 8% -Normandy (milk): 71% (udder): 24%
20 15 10
5 0
.018
Few male progeny tested each year in French dairy goats Progeny tested buck each year 40 Alpine
30 Saanen
100 daughters /year
with performances
25 Alpine
CDasc : 38%
AI bucks
> 5.5 years Sire-son
20% AI
15 Saanen
< 4 years Sire-daughter Alpine or Saanen
2011: Illumina goat SNP50 Bead Chip .019
Alpine and Saanen : two different breeds according to LD persistence between both 1 Manech red faced - Manech black faced Manech red faced - Lacaune Alpine - Saanen
0,9
Correlation of r
0,8 0,7
Other species : 0.9 Jersey-Holstein (70kb) 0.3 Jersey-Holstein (600kb) 0.2 Charolais-Angus (600kb)
0,6 0,58 0,5 0,4 0,3 0,2
0,08
0,1
0 0 70
500 600
1000 Marker distance (kb)
1500
2000
.020
