210 likes | 228 Views
Cow Adjustments for Genomic Predictions of Holstein and Jersey Bulls. Beltsville Agricultural Research Center. Centennial • 1910-2010. Introduction. Most countries use only bull PTAs in genomic prediction equations
E N D
Cow Adjustments for Genomic Predictions of Holstein and Jersey Bulls
Beltsville Agricultural Research Center Centennial • 1910-2010 Introduction • Most countries use only bull PTAs in genomic prediction equations • Information from genotyped cows was not increasing reliability of yield traits • Inflated PTA values of cows cause genomic predictions to suffer in accuracy • Solution was to make cow contributions comparable to those from bulls
Beltsville Agricultural Research Center Centennial • 1910-2010 DGV vs Traditional PTA (Bulls) 907 680 454 Milk (kg) 226 0 -70 125 267 360 436 501 568 636 716 847 -226 PA Milk (kg) Bull DGV Bull Traditional PTA DGV – evaluation based on genomics only Traditional PTA – no genomics
Beltsville Agricultural Research Center Centennial • 1910-2010 DGV vs Traditional PTA (Cows) 1134 907 680 434 226 Milk (kg) 0 -324 129 261 360 435 501 568 641 722 878 -226 -454 PA Milk (kg) Cow DGV Cow Traditional PTA DGV – evaluation based on genomics only Traditional PTA – no genomics
1134 907 680 Cow Std. Dev of Dereg M.S. (Milk, kg) Bull 454 680 226 362 272 0 181 Cow 0.4 0.6 1.0 2.5 Milk (kg) Bull 91 Daughter Equivalent (progeny) 0 -91 2000 2001 2002 2003 2004 2005 2006 2007 -181 Birth year Adjustments Needed Variance Adjustment Mean Adjustment
Beltsville Agricultural Research Center Centennial • 1910-2010 Cow Adjustment • Deregressed Mendelian Sampling (MS) = (PTA-PA) / f(REL) • Adj. MS = .84*MS - 784 • Adj. PTA = f(REL)*(Adj. MS+ PAn) + (1- f(REL)*PAn) f(REL) = weight in PTA from own records and progeny
DGV vs Traditional PTA (Cows) 1134 907 680 434 226 Milk (kg) 0 -324 129 261 360 435 501 568 641 722 878 -226 -454 PA Milk (kg) Cow DGV Cow Traditional PTA Adjusted Traditional PTA DGV – evaluation based on genomics only Traditional PTA – no genomics
Beltsville Agricultural Research Center Centennial • 1910-2010 Validation Populations • Predictor population - Animals with August 2006 evaluations • No Females • Unadjusted • Adjusted • Predicted population – Bulls with no evaluation in August 2006 but did have an evaluation in June 2010
Effects on Regression (ß) Beltsville Agricultural Research Center Centennial • 1910-2010 Deregressed value = α + ß·PTA
Interbull Validation of Regression Beltsville Agricultural Research Center Centennial • 1910-2010 Within 2 SE of expected to pass Fail Validation Test Pass Validation Test
Effects on Bias Beltsville Agricultural Research Center Centennial • 1910-2010 Bias = actual - predicted
Effects on Genomic Reliability Beltsville Agricultural Research Center Centennial • 1910-2010
Unadjusted Protein SNP effects Abs (SNP Effect)
Unadjusted Protein SNP effects (PAR) Abs (SNP Effect)
Adjusted Protein SNP effects (PAR) Abs (SNP Effect)
Adjusted Protein SNP effects (PAR) Abs (SNP Effect)
Beltsville Agricultural Research Center Centennial • 1910-2010 Benefit of Adjustment • Regressions closer to 1 • Reduction in Bias • Gain in Genomic Reliability • SNP estimates less affected by sex • Similar benefits for Jersey
Beltsville Agricultural Research Center Centennial • 1910-2010 The future • Investigate solutions to the problem of not being able to compare genotyped and non-genotyped cows • Reduce heritability • Add dam-herd interaction • Varying heritability by herd • Vary adjustments by sub-population • Increased genotyping with 3K chip will change genotyped population, which may necessitate modification of adjustment
Thank You! • Dr. John Cole – SNP effect graphics • AIPL Staff