analysis of genomic predictor population
DESCRIPTION
Analysis of Genomic Predictor Population. Objectives. E valuate the accuracy of cow and bull traditional information in the genomic evaluation system for Holstein How useful are cows? Would more old bulls increase accuracy? What is the contribution of additional predictor animals?. - PowerPoint PPT PresentationTRANSCRIPT
Cooper, 2014CDCB Meeting Aug. 5(1)
T.A. Cooper, G.R. Wiggans and P.M. VanRadenAnimal Genomics and Improvement Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705-2350
Analysis of Genomic Predictor Population
Cooper, 2014CDCB Meeting Aug. 5(2)
Objectives
Evaluate the accuracy of cow and bull traditional information in the genomic evaluation system for Holstein
How useful are cows?
Would more old bulls increase accuracy?
What is the contribution of additional predictor animals?
Cooper, 2014CDCB Meeting Aug. 5(3)
Introduction
The number of females genotyped monthly has increased from approximately 1,800 per month in 2010 to 12,650 per month in 2013.
Only a few countries other than the US include cows in the reference population. Ex: Ireland, Australia and Czeck Republic
For bulls, efforts have been made to increase collaborations with many other organizations.
Cooper, 2014CDCB Meeting Aug. 5(4)
Number of Genotypes Added Monthly
Aug-09
Feb-1
0
Jun-10
Sep-1
0
Dec-1
0
Mar-1
1
Jun-11
Sep-1
1
Dec-1
1
Mar-1
2
Jun-12
Sep-1
2
Dec-1
2
Mar-1
3
Jun-13
Sep-1
3
Dec-1
3
Mar-1
4
Jun-14
0
5000
10000
15000
20000
25000
30000
Males Females
No.
of
new
gen
otyp
es
Cooper, 2014CDCB Meeting Aug. 5(5)
Cut off study (Bulls vs. Cows)
Predictor Bulls – 21,883
Predictor Cows – 30,852
Traditional evaluation by August 2012 to predict animals that gained a traditional evaluation between August 2012 and December 2013
Only females who were genotyped before they were two years old where included in the validation set to avoid selection bias
Cooper, 2014CDCB Meeting Aug. 5(6)
Bulls and/or Cows Predicting Bulls Gain Gen Rel
TraitValidation Bulls (no.)
Parent Average
Cows Only
Bulls Only Both
Milk 1486 40.5 26.2 34.9 35.8Fat 1486 40.5 21.9 33.9 33.3Protein 1486 40.5 16.8 25.8 26.0PL 1484 35.3 21.4 54.9 55.2 SCS 1484 37.5 16.9 30.3 31.1DPR 1287 34.9 −5.0 23.3 22.6HCR 1377 28.2 −3.2 28.3 25.1CCR 1063 28.6 19.5 56.9 57.4Final score 934 36.3 17.7 27.0 26.0Average* 20.4 31.7 32.1
Highest Gen Rel Gain *Excluding $NM
Cooper, 2014CDCB Meeting Aug. 5(7)
Gain Gen Rel
TraitValidation Bulls (no.)
Parent Average
Cows Only
Bulls Only Both
Milk 26559 26.1 23.9 30.2 31.8Fat 26562 26.1 20.2 26.4 27.5Protein 26558 26.1 14.1 18.8 20.1PL 357 19 −6.4 8.5 14.6SCS 24879 24.8 15.5 25.1 26.0DPR 9064 24.6 −1.1 −0.8 −0.6HCR 12618 24.5 1.6 9.3 13.1CCR 7156 24.4 25.4 8.6 15.2Final score 10759 21.9 19.5 8.3 29.8Average* 20.6 23.0 25.5
Bulls and/or Cows Predicting Cows
Highest Gen Rel Gain *Excluding $NM
Cooper, 2014CDCB Meeting Aug. 5(8)
Number predictor bulls by birth year
<19851985
19861987
19881989
19901991
19921993
19941995
19961997
19981999
20002001
20022003
20042005
20062007
20082010
0
500
1000
1500
2000
2500
Birthyear
No.
bu
lls
wit
h t
rad
itio
nal
eva
luat
ion
s
Cooper, 2014CDCB Meeting Aug. 5(9)
Cut off study (Age)
Predictor Bulls
− All – 21,883− Bulls born before 1996 removed – 17,047− Bulls born before 2001 removed – 11,507− Bulls born before 2005 removed – 6,623
Traditional evaluation by August 2012 to predict animals that gained a traditional evaluation between August 2012 and December 2013
Cooper, 2014CDCB Meeting Aug. 5(10)
Bulls only excluded by birth year
Gain Gen Rel
TraitValidation Bulls (no.)
Parent Average All
Birth years included≥1996 ≥2001 ≥2005
Milk 1486 40.5 34.9 34.7 33.5 30.2Fat 1486 40.5 33.9 34.1 32.7 28.7Protein 1486 40.5 25.8 26.3 25.3 21.9PL 1486 35.3 54.9 53.0 48.9 36.6SCS 1484 37.5 30.3 29.3 26.6 23.3DPR 1287 34.9 23.3 21.9 19.5 3.7HCR 1377 28.2 28.3 24.0 23.5 10.8CCR 1063 28.6 56.9 35.1 30.0 5.5Final score 934 36.3 27.0 25.6 22.8 15.9Average* 31.7 29.9 27.4 19.7
Highest Gen Rel Gain *Excluding $NM
Cooper, 2014CDCB Meeting Aug. 5(11)
All 1996 (25%) 2001 (50%) 2005 (75%)15
20
25
30
35
40
Random removal of bulls (%) Bulls only in predictor set
Birth year removal of bulls (yr) Bulls only in predictor set
Birth year removal of bulls (yr) Bulls and Cows in predictor set
Random vs. birth year exclusion (Milk)G
ain
in g
enom
ic r
el (
Mil
k)
Cooper, 2014CDCB Meeting Aug. 5(12)
Conclusions
How useful are cows?
− Cows contribute a small amount to genomic accuracy due to low reliabilities in the US. However, if bull genotypes are limited, they become more valuable.
Would more old bulls increase accuracy?
− Historic bulls contribute a small amount to genomic accuracy due to linkage decay. Bulls closer in age to the young bulls offer more predictive ability.
What is the contribution of additional predictor animals?
− We have not yet reached a plateau of gains in genomic reliability.
Cooper, 2014CDCB Meeting Aug. 5(13)
Questions?
Holstein and Jersey crossbreds graze on American Farm Land Trust’sCove Mountain Farm in south-central PennsylvaniaSource: ARS Image Gallery, image #K8587-14; photo by Bob Nichols
AIP web site:http://aipl.arsusda.gov