Genomic selection for crossbred performance accounting for breed-specific effects

Table 4 Prediction accuracy of performance of crossbred sows for gestation length and litter size

Model^a	Training^b	Accuracy^c	SD^d
Litter size
GS	LW	0.06	0.10
	LR	0.07	0.11
	F1	0.23	0.08
BS	LW	0.06	0.11
	LR	0.06	0.13
	LW and LR*	0.09	0.12
	F1_LW	0.21	0.08
	F1_LR	0.12	0.09
	F1_LW and F1_LR*	0.23	0.08
Gestation length
GS	LW	0.42	0.08
	LR	0.30	0.09
	F1	0.52	0.08
BS	LW	0.39	0.08
	LR	0.23	0.10
	LW and LR*	0.45	0.08
	F1_LW	0.43	0.08
	F1_LR	0.34	0.09
	F1_LW and F1_LR*	0.53	0.08

* Predicted direct genomic value was the “total direct genomic value” (sum of the breed-specific direct genomic values)
^aGS, traditional genomic selection model; BS, model that accounts for breed-specific effects
^bLW, Large White; LR, Landrace; F1, two-way crossbred; F1_LW, alleles of the F1 population inherited from the LW population; F1_LR, alleles of the F1 population inherited from the LR population. Training populations in each replicate were defined as a random set of 90% (N = 832) of the animals used for the estimation of variance components
^cAverage of the 40 replicates; accuracy was defined as the correlation between the direct genomic values of the validation population [random set of 10% (N = 92) of the crossbred animals used for the estimation of variance components] and their average pre-adjusted phenotypes in each replicate
^dStandard deviation over replicates; the highest accuracies for each model and trait are indicated in bold

ISSN: 1297-9686