Simultaneous fitting of genomic-BLUP and Bayes-C components in a genomic prediction model

Table 3 Accuracy (SE) of the predicted values for the youngest sires based on the different prediction methods

Trait_(π)	G-BLUP	Bayes-C	GBC
SCC _{(20%, 20%,)}	0.602 (0.066)	0.604 (0.064)	0.607 (0.065)
Fkg _{(10%, 10%,)}	0.716 (0.049)	0.733 (0.042)	0.731 (0.047)
Mkg _{(10%, 10%,)}	0.705 (0.051)	0.701 (0.050)	0.719 (0.048)
Pkg _{(10%, 1%,)}	0.695 (0.053)	0.689 (0.050)	0.696 (0.051)
Average	0.679	0.682	0.688

\({\text{Accuracy }} = \frac{{corr\left( {DYD,GEBV} \right)}}{{\sqrt {r_{DYD}^{2} } }}\)
SE: standard errors computed from 10,000 bootstrap samples
G-BLUP: genomic BLUP using genomic-based relationship matrix; Bayes-C: a non-linear method that fits zero effects and normal distributions of effects for SNPs; GBC: an iterative method that fits a G-BLUP next to SNP effects with a Bayes-C prior
SCC, somatic cell count; Fkg, fat yield; Mkg, milk yield; Pkg, protein yield
π refers to the optimal π values (i.e. proportion of SNP having large effects) when using Bayes-C and GBC

ISSN: 1297-9686