27.03.2013 Views

Genomic Selection in Dairy Cattle

Genomic Selection in Dairy Cattle

Genomic Selection in Dairy Cattle

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>Genomic</strong> <strong>Selection</strong> <strong>in</strong><br />

<strong>Dairy</strong> <strong>Cattle</strong><br />

AQUAGENOME<br />

Applied Tra<strong>in</strong><strong>in</strong>g Workshop, Sterl<strong>in</strong>g<br />

Hans Daetwyler, The Rosl<strong>in</strong> Institute and R(D)SVS


• <strong>Dairy</strong> <strong>in</strong>troduction<br />

Overview<br />

• TTraditional diti l bbreed<strong>in</strong>g di<br />

• <strong>Genomic</strong> selection<br />

• Advantages<br />

• Disadvantages<br />

• <strong>Genomic</strong> selection around the world<br />

• SSummary mmar


<strong>Dairy</strong> <strong>Cattle</strong><br />

• Most are specialised dairy breeds<br />

– Holste<strong>in</strong>, Jersey, Brown Swiss, Ayrshire, etc<br />

• Some dual purpose<br />

– Simmental, Montbeliarde,<br />

NNormande, d etc<br />

t


<strong>Dairy</strong> Industry<br />

• Relies very heavily (>80%) on artificial<br />

<strong>in</strong>sem<strong>in</strong>ation (AI)<br />

• Nucleus herds exist but a large part of<br />

bbreed<strong>in</strong>g di iis still ill ddone bby ffarmer bbreeders d<br />

• Global exchange of genetic material<br />

between countries and AI companies<br />

– International genetic g<br />

evaluation


• Production<br />

Traits selected for<br />

– Milk, fat and prote<strong>in</strong> yield...<br />

• Conformation<br />

– Udder, legs, capacity...<br />

• Functional & Health<br />

– Somatic cell score, fertility,....


Cornerstone of genetic g pprogress g<br />

• AI Companies <strong>in</strong>crease the accuracy of<br />

selection through progeny test<strong>in</strong>g.<br />

– Young bulls sire approx approx. 100 daughters which<br />

provide <strong>in</strong>formation for their EBV once the<br />

daughters produce records<br />

– Only 1 <strong>in</strong> 10 (or less) of bulls return to active<br />

service<br />

– Very costly


Bull<br />

Life cycles<br />

Birth 1 yr 2 yr 5 yr 10 yr<br />

Selected on parent avg<br />

Cow<br />

Daughters g born<br />

Bull is<br />

Son is progeny<br />

progeny tested tested<br />

Birth 1 yr 2 yr 4 yr 7 yr<br />

Gives Birth<br />

Daughter Gives<br />

Birth<br />

Son is progeny<br />

tested


Pedigree +<br />

Records<br />

Traditional Breed<strong>in</strong>g<br />

Estimation<br />

• Accuracy <strong>in</strong>creased with:<br />

• Own phenotypic records<br />

• Information on relatives (sibs, progeny,…)<br />

Breed<strong>in</strong>g<br />

Vl Value


• Works well <strong>in</strong>:<br />

Traditional Breed<strong>in</strong>g<br />

– Med.- high heritability traits<br />

– Own records or progeny data is available<br />

• Less effective<br />

– Low heritability traits, sex limited traits<br />

– Young animals (no records, no progeny)


Genotypic Information<br />

• Thousands of s<strong>in</strong>gle nucleotide<br />

polymorphisms (SNP) are now available <strong>in</strong><br />

many species i<br />

• SNP have 2 alleles or gene variants<br />

1 2 2 1 2 1 2<br />

2 1 2 1 1 2 2


Genotypic Information<br />

• Thousands of s<strong>in</strong>gle nucleotide<br />

polymorphisms y (SNP) ( ) are now available <strong>in</strong><br />

many species<br />

1 2 2 1 2 1 2<br />

2 1 2 1 1 2 2<br />

Recode 1 1 0 2 1 1 0


Genotype y Data for<br />

Elevation - Chromosome 1<br />

1000111220020012111011112111101111001121100020122002220111<br />

1202101200211122110021112001111001011011010220011002201101<br />

1200201101020222121122102010011100011220221222112021120120<br />

2010020220200002110001120201122111211102201111000021220200<br />

0221012020002211220111012100111211102112110020102100022000<br />

2201000201100002202211022112101121110122220012112122200200<br />

0200202020122211002222222002212111121002111120011011101120<br />

0202220001112011010211121211102022100211201211001111102111<br />

2110211122000101101110202200221110102011121111011202102102<br />

1211011022122001211011211012022011002220021002110001110021<br />

1021101110002220020221212110002220102002222121221121112002<br />

0110202001222222112212021211210110012110110200220002001002<br />

0001111011001211021212111201010121202210101011111021102112<br />

2111111212111210110120011111021111011111220121012121101022<br />

202021211222120222002121210121210201100111222121101<br />

• From Filippo Miglior, Canadian <strong>Dairy</strong> Network.


The Opportunity<br />

• Genotyp<strong>in</strong>g gives us ‘picture/snapshot’ of<br />

the genetic makeup of an animal<br />

• The more SNP the clearer the picture (up to a limit)<br />

• This new source of <strong>in</strong>formation can now<br />

(or soon) be used <strong>in</strong> genetic evaluation by:<br />

– Comb<strong>in</strong><strong>in</strong>g genotyp<strong>in</strong>g data with traditional<br />

pedigree and phenotypic records


Method<br />

• Meuwissen et al., Genetics, 2001<br />

– Divide the genome g <strong>in</strong>to many y segments g ( (each<br />

with 1+ markers)<br />

– Estimate the genetic g effect of each segment g<br />

from a sample of <strong>in</strong>dividuals<br />

• Many small black boxes<br />

– Genotype another population sample and<br />

sum the segment effects to get a breed<strong>in</strong>g<br />

value l ffor eachh


How is this different?<br />

Pedigree +<br />

Records +<br />

<strong>Genomic</strong> Data<br />

Estimate SNP<br />

effects<br />

<strong>Genomic</strong><br />

Breed<strong>in</strong>gg<br />

Value


How is this different?<br />

Pedigree +<br />

Records +<br />

<strong>Genomic</strong> Data<br />

Estimate SNP<br />

effects<br />

<strong>Genomic</strong> Breed<strong>in</strong>g<br />

Values for animals<br />

WITHOUT records!<br />

<strong>Genomic</strong><br />

Breed<strong>in</strong>gg<br />

Value


2 Ma<strong>in</strong> <strong>Genomic</strong> Estimation Methods<br />

• <strong>Genomic</strong> BLUP<br />

– Easy to implement because similar to classic<br />

BLUP<br />

• Bayesian methods<br />

– More complicated<br />

– May have higher accuracy than GBLUP<br />

• Papers say yes, yes practice says not so much much…


Advantages of <strong>Genomic</strong> <strong>Selection</strong><br />

• Increase genetic ga<strong>in</strong><br />

– BBy <strong>in</strong>creas<strong>in</strong>g i i accuracy of f selection l ti<br />

• Parent avg 40%, <strong>Genomic</strong> BVs higher<br />

– BBy reduc<strong>in</strong>g d i th the generation ti i<strong>in</strong>terval t l<br />

• Select animals before they are of productive and/or<br />

reproductive age<br />

• Reduce/elim<strong>in</strong>ate the need for progeny test<strong>in</strong>g<br />

– Reduces cost


Bull<br />

Life cycles<br />

Birth 1 yr 2 yr 5 yr 10 yr<br />

Selected on parent avg<br />

Cow<br />

Daughters g born<br />

Bull is<br />

Son is progeny<br />

progeny tested tested<br />

Birth 1 yr 2 yr 4 yr 7 yr<br />

Gives Birth<br />

Daughter Gives<br />

Birth<br />

Son is progeny<br />

tested


Life cycles<br />

Bull<br />

Select Select Select Great- Great<br />

Sons here Grand-sons grand-sons<br />

Birth 1 yr 2 yr 5 yr 10 yr<br />

Selected on parent avg<br />

Cow<br />

Daughters g born<br />

Bull is<br />

Son is progeny<br />

progeny tested tested<br />

Birth 1 yr 2 yr 4 yr 7 yr<br />

Gives Birth<br />

Daughter Gives<br />

Birth<br />

Son is progeny<br />

tested


Advantages<br />

• Lower rate of <strong>in</strong>breed<strong>in</strong>g per generation<br />

(Daetwyler ( y et al., JABG, 2007) )<br />

– Moves from family selection to <strong>in</strong>dividual selection<br />

– Example:<br />

• Parent average same for full sib newborns<br />

• <strong>Genomic</strong> BV different for full sibs<br />

– If generation <strong>in</strong>tervals are shortened substantially<br />

then annual <strong>in</strong>breed<strong>in</strong>g rates could be higher


Inbreed<strong>in</strong>g rates of methods<br />

Inbreed<strong>in</strong>g<br />

ratee<br />

per gen. (%)<br />

3<br />

2<br />

1<br />

0<br />

Phenotypic Sel.<br />

Classic BLUP<br />

<strong>Genomic</strong> Sel.<br />

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1<br />

Heritability<br />

Daetwyler et al., J. Anim. Breed. Genet., 2007


Advantages g<br />

• Once marker effects are estimated they<br />

can bbe used d ffor a ffew generations ti<br />

– BUT accuracy will reduce <strong>in</strong> each generation<br />

if not t re-estimated ti t d<br />

• <strong>Selection</strong> on novel traits, expensive<br />

phenotyp<strong>in</strong>g<br />

• New breed<strong>in</strong>g g strategies g


Disadvantages<br />

• New method, not fully proven and tested<br />

• Need to genotype a sufficiently large set of<br />

animals for accurate marker estimates<br />

(Daetwyler et al., PLoS One, 2008)<br />

• Lower heritability more records needed<br />

• Marker estimates must be estimated <strong>in</strong><br />

population that they will be used <strong>in</strong><br />

• Across breed accuracy low


Disadvantages<br />

• Genotypes still costly (125£ dairy)<br />

• Some species have no dense marker<br />

maps yet<br />

• When generation <strong>in</strong>tervals are already low<br />

genetic ga<strong>in</strong> due to genomic selection will<br />

be less<br />

• In large litters accuracy can be ga<strong>in</strong>ed<br />

ffrom <strong>in</strong>fo i f on sibs ib lless advantage d t of f GS


Genotyp<strong>in</strong>g<br />

• Illum<strong>in</strong>a Bov<strong>in</strong>eSNP50 TM BeadChip<br />

– 58,000 genetic markers, equally spaced<br />

– 38,416 used <strong>in</strong> genomic predictions<br />

• Total of one-third non-<strong>in</strong>formative (currently) or<br />

without variation across dairy cattle<br />

– Openly commercialized to several<br />

laboratories <strong>in</strong> various countries<br />

– Current genotyp<strong>in</strong>g cost ≈$250 USD per<br />

animal<br />

Miglior, Can. <strong>Dairy</strong> Network


<strong>Genomic</strong> selection around the<br />

world ld<br />

• USA & Canada (N.A.) Collaboration<br />

• New Zealand (LIC)<br />

• Netherlands (CRV)<br />

• Australia (ADHIS & co.)<br />

• DDenmark k & SSweden d (Viki (Vik<strong>in</strong>g GGenetics) ti )<br />

• Several other countries likely to follow<br />

fairly soon<br />

Miglior, Can. <strong>Dairy</strong> Network


Trait<br />

Australian results<br />

Table 1. Accuracy of genomic breed<strong>in</strong>g values calculated at time of birth for<br />

Genetic Australia’s 2003 progeny test team with two genomic selection<br />

methods, BLUP and a Bayesian method (BAYES).<br />

Records <strong>in</strong><br />

reference<br />

population<br />

Number of<br />

SNPs used<br />

Sire pathway<br />

EBV GBLUP GBayesA<br />

AUS <strong>Selection</strong><br />

Index 637 3889 0.62 0.66 0.69<br />

AUS Profit<br />

Rank<strong>in</strong>g 635 3414 0.59 0.73 0.74<br />

Prote<strong>in</strong> yield 637 4055 0.53 0.67 0.69<br />

Prote<strong>in</strong> % 637 4369 0.45 0.54 0.60<br />

Fertility 332 3090 0.40 0.42 0.37<br />

Hayes et al, Review, JDS, 2008


• North America<br />

World wide results<br />

– GS Accuracy avg 71%, PA 52%<br />

– Diff Difference bbetween t BLUP and d BBayesian i was 1%<br />

• Netherlands<br />

– Large <strong>in</strong>creases <strong>in</strong> accuracy of GS over PA<br />

– Use Bayesian y methods<br />

• New Zealand<br />

– GS accuracy avg 70 – 80%, PA 58%<br />

Hayes et al, Review, JDS, 2008


Higher risk vs. Lower risk<br />

• How will different AI companies use it?<br />

• Examples:<br />

– LIC: Reduce progeny test<strong>in</strong>g from 300 to 100-<br />

150 bulls, emphasis on DNA teams<br />

– Recent recall announcement for HO team(s)<br />

– CRV: Reduce progeny p g y test<strong>in</strong>g g from 500 to<br />

300 bulls, reduce YS <strong>in</strong>centives<br />

– North-American units likely y to use a more<br />

conservative transitional approach<br />

Miglior, Can. <strong>Dairy</strong> Network


Possible strategy<br />

• Genotype a large number of elite females<br />

and bull calves<br />

• Put the best GEBV bulls <strong>in</strong>to organized<br />

progeny test<strong>in</strong>g<br />

• Use the best of those as sires of sons, and<br />

<strong>in</strong> teams for the GEBV bull market<br />

• Use the best proven bulls for the proven<br />

bull bu market a et<br />

Miglior, Can. <strong>Dairy</strong> Network


How can we do better?<br />

• Genotype more SNP to get clearer<br />

‘picture’ picture of genetic variation (up to a limit)<br />

• Genotype and get records for more<br />

animals i l<br />

• Ref<strong>in</strong>e estimation methods<br />

– Determ<strong>in</strong>e when to use BLUP or Bayes<br />

• Develop p<br />

new estimation methods


Summary<br />

• <strong>Dairy</strong> <strong>in</strong>dustry uniquely suited for genomic<br />

selection<br />

– Faster genetic progress possible with higher<br />

accuracy and shorter generation <strong>in</strong>tervals<br />

• Several countries are implement<strong>in</strong>g<br />

genomic selection<br />

– Hybrid systems merg<strong>in</strong>g classic and genomic<br />

selection


Summary<br />

• UUsefulness f l of f genomic i selection l ti ddepends d<br />

on:<br />

– PPopulation l ti structure/history<br />

t t /hi t<br />

• Size of sib families<br />

• Generation <strong>in</strong>terval<br />

– Availability y of dense marker maps p<br />

– Availability of many genotyped <strong>in</strong>dividuals<br />

with records


Acknowledgements<br />

• Filippo Miglior, Canadian <strong>Dairy</strong> Network,<br />

Guelph<br />

• BBen HHayes, Vi Victoria t i DDep. PPrimary i<br />

Industries, Melbourne<br />

• And my y fund<strong>in</strong>g... g<br />

SABRETRAIN is funded by the Marie Curie Host fellowships for<br />

Early Stage Research Tra<strong>in</strong><strong>in</strong>g fund<strong>in</strong>g mechanism, as part of the<br />

6th Framework Programme of the European Union European<br />

Commission.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!