Single-step genomic BLUP for national beef cattle evaluation in US: from initial developments to final implementation Daniela Lourenco S. Tsuruta, B.O. Fragomeni, Y. Masuda, I. Aguilar A. Legarra, S. Miller, D. Moser, I. Misztal 11 th WCGALP 2018
Angus Main beef cattle breed in USA Genomic Selection since 2009 2
Multistep Genomic Evaluation Records for Calibration 108,211 38,988 57,550 2,253 11,756 2010 2012 2013 2014 2016 Kachman, 2008 3
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Problems with Multistep Big fluctuations in GEBV for new calibration Rank change for bulls with high accuracy Overfitted models 2x the number of traits High genetic correlation between phenotype and MBV 0.8 Marbling 0.6 0.4 0.2 0-0.2-0.4 Multistep Traditional 4
Single-step genomic BLUP (ssgblup) Pedigree Phenotypes SNP ssgblup H 1 =A 1 + 0 0 0 G 1 A 1 22 Aguilar et al., 2010 GEBV UGA group (2008 now) 5
Initial tests of ssgblup for Angus Number of Genotyped Animals 406,033 442,635 303,246 335,325 219,849 184,354 82,000 112,000 132,000 152,000 07 2014 01 2015 07 2015 10 2015 01 2016 07 2016 02 2017 05 2017 11 2017 01 2018 6
Ability to predict future performance 2014 8M animals in pedigree 6M BW and WW 3.4M PWG 52k genotyped animals 18.7k born in 2013 2017 10M animals in pedigree 8M BW and WW 4.2M PWG 335k genotyped animals 18.7k born in 2016 Predictive ability direct = COR(Y_adj, GEBV) Predictive ability maternal = COR(Y_adj, total_maternal_gebv) 7
Ability to predict future performance Predictive Ability Average Gain 0.47 Direct 0.39 0.38 0.34 0.40 0.35 2014 = 25% 0.29 0.29 2017 = 36% 0.23 Maternal 2014 = 8% 2017 = 10% BW WW PWG BLUP ssgblup14 ssgblup17 8
USMARC comparisons of ssgblup x multistep USMARC Predictive Ability Correlation with MARC EBV for 143 bulls 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 BW WW YWT 0 BW WW YWT MILK CWT MARB REA FAT MS SS MS SS Kuehn et al., 2017 9
1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 Genetic trends for carcass traits Marbling Carcass Weight 0.7 50 0.6 0.5 40 0.4 30 0.3 20 0.2 0.1 10 0 0-0.1-0.2-10 -0.3-20 ssgblup Multistep Traditional ssgblup Multistep Traditional 10
Increasing number of genotyped animals Number of genotyped animals increased 5-fold from 2014 to 2018 150,000 > 2 hours > 700Gb RAM APY ssgblup Borrowed from algorithm to construct A -1 Core and Non-core 1 = G 1 cc 0 0 0 + G 1 cc Gcn I G APY Mnn 1 GncGcc 1 I Mnn = g ii g ic Gcc 1 g ci Misztal et al., 2014 11
APY ssgblup in 2014 core APY G -1 G -1 non-core How to choose core animals? 12
Correlation (GEBV,GEBV_apy) Correlation (GEBV,GEBV_apy) APY ssgblup in 2014 PWG Core based on accuracy 0.99 0.99 0.99 PWG Random Core 0.99 0.99 0.99 0.97 0.97 0.94 28 min 16 Gb 2k 4k 8k 10k 33k Lourenco et al., 2015a Lourenco et al., 2015b 5k 10k 15k 20k Regular inversion = 213 min 230 Gb 13
Number of Eigenvalues COR (GEBV,GEBV_APY) How to choose the number of core in APY? Ne, Me, ESM, Eigen of G Limited dimensionality Pocrnic et al., 2016 Misztal, 2016 AAA 82k 14555 1.00 AAA - 82k 3654 6166 10605 0.98 0.96 0.94 6166, 0.98 10605, 0.99 14555, 0.99 90 95 98 99 % of Variance 0.92 0.90 3654, 0.96 0 2000 4000 6000 8000 10000 12000 14000 16000 NUMBER OF EIGENVALUES 14
Additional features in ssgblup Commercial products e.g. GeneMax for non-registered animals Based on SNP effects Accurate SNP effects with APY? 15
SNP effects in APY ssgblup a G = λd Z G 1 u G 1 a G 1 Z a a 1 GAPY = λd Z 1 G APY u APY 1 G APY 1 G APY a Gcc 1 = λd Z G 1 CC u APY G 1 cc G 1 cc 16
Additional features in ssgblup Interim evaluations Indirect predictions Quick evaluations between official runs Should be comparable to GEBV 17
Indirect predictions for young animals W W + α A 1 + α 0 0 0 G 1 A 1 u= W y 22 GEBV y = w 1 PA + w 4 DGV w 5 PP GEBV y DGV parent average yield deviation progeny contribution direct genomic value pedigree prediction GEBV = w 1 PA + w 2 YD + w 3 PC + w 4 DGV w 5 PP Lourenco et al., 2015 18
Problem with Indirect predictions COR(GEBV,DGV) > 0.99 Lourenco et al., 2015 Avg(GEBV) 100 Avg(DGV) 0 Base of SSGBLUP: modelled as a mean in genotyped animals p u g = N 1μ, G Vitezica et al. (2011) μ = (Pedigree base) (Genomic base) 19
Correcting for bias of indirect predictions 120 100 80 60 40 20 0-20 GEBV u ip Za Legarra_2017 Double_Fit Average_GEBV E u a = μ + Z a DGV = GEBV + Z a Lourenco et al., 2018 20
Issues in the implementation of ssgblup for Angus 1) Omega = 0.7 indicates inflation in GEBV Inbreeding NO Inbreeding Inbreeding Inbreeding Solution: adding inbreeding for A -1 removed inflation in GEBV Omega = 1.0 21
Issues in the implementation of ssgblup for Angus 2) Inclusion of external EBV into growth evaluation 10k Red Angus EBV External EBV + genomics was not supported E = external I = internal T = PEV for E Adapted from Legarra et al., 2007 22
Issues in the implementation of ssgblup for Angus 3) Calving ease evaluation was not quite easy BW + CE in linear-threshold model BLUP = 12 hours 152k genotyped animals APY ssgblup = 4.5 days Scenario Description of parameters correlation with rounds hours pcg rounds alpha beta genomic traditional 40-60 12 - genomic 40 0.9 0.1 488 108-1 100 0.9 0.1 81 43 0.999 2 100 0.85 0.15 62 32 0.999 3 200 0.9 0.1 24 25 0.999 4 200 0.85 0.15 19 19 0.999 23
Issues in the implementation of ssgblup for Angus 4) Accuracy of GEBV Large datasets Impossible to invert d i r and d i p are approximated Accuracy = 1 - LHS -1 (Misztal and Wiggans, 1988) Diag(C ZZ+ ) = PEV ii LHS uu = 1 (λ + di r + d i p ) 24
Issues in the implementation of ssgblup for Angus 4) Accuracy of GEBV Z Z+ λa 1 + λ 0 0 0 G 1 A 1 22 LHSii uu = 1 (λ + di r + d i p + d i g ) d i r d i p d i g = var_ratio *[Rel + (1 g ii ) + zeta Rel Rel PA ] 25
Issues in the implementation of ssgblup for Angus 4) Accuracy of GEBV Cor = 0.87 Avg_True = 0.55 Avg_approx. = 0.50 MSE = 0.0035 26
Implementation of ssgblup on 7/7/2017 Current Angus evaluation with ~ 450k 19k core Weekly evaluations ~ 18 traits (maternal, categorical, external information) Indirect predictions based on SNP effects a Gcc 1 = λd Z G 1 CC u APY Minimal changes for proven animals Considerable changes for young animals More variation among half- and full-sibs 27
Final Remarks ssgblup tests were extensive and took couple of years More stable than multistep Implementation of ssgblup by Angus raised several issues All solved Successful weekly evaluations for 7 months Evaluation with ~450k genotyped animals is possible with APY Implementation of ssgblup for Angus in 2017 set new standards for beef cattle evaluation in USA 28
Acknowledgements 29