Application of MAS in French dairy cattle. Guillaume F., Fritz S., Boichard D., Druet T.

Size: px
Start display at page:

Download "Application of MAS in French dairy cattle. Guillaume F., Fritz S., Boichard D., Druet T."


1 Application of MAS in French dairy cattle Guillaume F., Fritz S., Boichard D., Druet T.

2 Considerations about dairy cattle Most traits of interest are sex linked Generation interval are long Recent emphasis on low heritability traits AI is widespread => number of key animal reduced Big amount of phenotypes and pedigree are available Bull sire Bull dams Proven bulls Marker Assisted Selection could be helpfull Young males Heifers First crop daughters

3 1 st step : QTL Detection Grand-Daugther design 1548 Bulls (14 sires families, 3 breeds) 169 microsatellites 24 traits Detection by within-sire linear regression Results : 32 QTL Contribution to genetic variance 6-40 % Confidence intervals ~20 cm

4 2 nd step : Building a MAS program Three breeds QTLs and markers: 45 markers 14 regions 2-5 markers / QTL 8 traits Milk production traits, SCC, fertility, udder depth Impact on breeding goals Numbers of QTL Proportion of variance explained This parameters have changed SCC, fertility, udder depth (2004) Markers changed in 2004 Available data are increasing In a first step key animal have been genotyped 10,000 new genotype/year (100,000 available in 2008)

5 Results : Linkage equilibrium Simulations Field data Linkage disequilibrium Simulations

6 Simulations study Analysis of MAS results when the true genetic values are known Large samples Available rapidly Simulations based on MAS data (2004 & 2006) Simulations based on MAS data (2004 & 2006) Same pedigree Same probability of identity by descent (informativity) Same animals with records Same weights for records

7 Simulations study Every QTL considered as bi-allelic Additive effects Parameters based on previous studies results % heterozygous in population 10-40% Milk Fat yield VAR (QTL) 5%-50 % Protein yield Fat % Prot % Nb QTL %VAR(G) 40 % 50 % 35 % 60 % 50 % 100 repetitions of the simulation

8 Evaluations Two models for evaluation : Classical evaluation (polygenic) Based on pedigree only MAS evaluation Based on pedigree and marker information Analysis based on 2 sets of candidates 1180 males in males in 2006 Computation of correlations : True breeding Value x

9 Increase of s reliability MAS- Difference MAS- Difference Milk Fat yield Prot. yield Fat % Prot % Figures expressed as REL=(correlation(BV;)) 2

10 Increase of s reliability 2004 Sires without genotyped daugther Sires with genotyped daughter MAS- GAS- MAS- GAS- Milk Fat yield Prot. yield Fat % Prot % Figures expressed as REL=(correlation(BV;)) 2

11 Full sibs analyses 142 fullsibs pairs (100 repetitions) Selection of the best individual based on MAS Comparison with true genetic values Results divided according to expected differences : When fullsibs received the same allele s=0 When fullsib #1 received the favourable allele s=1 When fullsib #2 received the favourable allele s=-1 Score= summation of s over all QTL Large score => large differences due to QTL

12 Traits: milk production H2 = 30 % 4QTL ; 40 % Var(G) Full sibs Selection 6000 MAS GAS Classes 0 frequency 46.0 Correct choice (% ) 49.4 Diff. -5 Correct choice (%) 54.0 Diff et + 4 &

13 Conclusion MAS- better predict true breeding value than pedigree based With data accumulation MAS- superiority over classical increase Strategies can be setup to increase MAS- accuracy MAS- take well into account difference when they exist!

14 Results on field data Correlations between DYD of January 2007 and classical or MAS based on April 2004 evaluation of 899 progeny tested Holstein bulls for production traits weighted by DYD reliability in 2007 H 2 =0.30 H 2 =0.50 N QTL %V(G) 4QTL 40% 5QTL 50% 5QTL 35% 4QTL 60% 4QTL 50% Milk Fat yield Protein yield Fat content Protein content MAS Difference Expectation

15 Results Preliminary results 899 Bulls, approximate BV Selection already occurs Marker set changed With experience genotyped animal changed So Better results should be obtained in the coming years

16 Others results Most of the QTL have been confirmed Some new QTL have been discovered Experience have been gained regarding data collection Important amount of genetic material is available

17 Evolutions of program Fine mapping of QTL regions Use of denser SNP map Addition of new traits / new QTL Integration of Linkage Desequilibrium in evaluation Better accuracy of is expected

18 Expected results Average correlations between true and estimated breeding value of young bulls for fat % with different haplotypes length. (100 simulations, N= 576) MAS (4 Markers) MAS (6 Markers) MAS (8 Markers) MAS (10 Markers) MAS LA GAS MEAN CORRELATION (TBV/) H 2 = QTL 60% V(G) H 2 = QTL 50% V(G)

19 Conclusion MAS works! MAS program benefit from industry involvement Industry can benefit from MAS program New technologies will lead to improvements in accuracy

20 Thank you for your attention