Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: March 5, 2012
Publication Date: June 28, 2012
Citation: Van Raden, P.M., Wright, J.R., Cooper, T.A. 2012. Adjustment of selection index coefficients and polygenic variance to improve regressions and reliability of genomic evaluations. Journal of Dairy Science. 95(Suppl. 2):446–447(abstr. 449). 2012.
In multi-step genomic evaluations, direct genomic values (DGV) are computed using either marker effects or genomic relationships among the genotyped animals, and information from non-genotyped ancestors is included later by selection index. The DGV, the traditional evaluation (EBV), and a subset breeding value (SBV) estimated using pedigree relationships among the genotyped animals are combined according to theoretical weights based on reliabilities of the 3 terms. In official yield trait evaluations of young Holstein bulls, the weights average 0.99 for DGV, 0.12 for EBV, and -0.11 for SBV. Alternative weights have been proposed by other countries to increase reliabilities and regressions of predicting future data from past. Most U.S. regressions were close to expected values and increased when some weight was removed from the DGV and added to either the EBV or SBV or both. Reliabilities changed little when the weight was added to SBV, but reliabilities decreased slightly if weight was added to the EBV. Maximum weights on DGV of 1.0, 0.9, and 0.8 were compared. For each 0.1 decrease, regressions increased by about 0.02. Regressions for a few traits were lower than expected, and limiting the DGV weight to 0.9 or 0.8 instead of the theoretical value of 1.0 helped bring the regressions into compliance with validation tests. Adjustments to polygenic variance also increased the regressions by about 0.02 for each 0.1 increase, but reliabilities were slightly reduced when compared to adjusting the selection index weights. Finding optimum percentages of polygenic variance required more computation and was less flexible than finding optimum weights, because recombining 3 terms in a final step is easier than re-estimating all marker effects. Conclusions are that index adjustments can help to pass genomic validation tests for some traits by removing small biases in regressions, but that the theoretical selection index weights currently in use are close to ideal.