Improved precision of QTL mapping using a nonlinear Bayesian method in a multi-breed population leads to greater accuracy of across-breed genomic predictions

Table 5 Accuracy and bias of across-breed genomic predictions for milk production traits

Ref datasets	Prediction method	Validation dataset	FY		MY		PY		F%		P%		Avg.
Ref datasets	Prediction method	Validation dataset	Acc.	Bias	Acc.	Bias	Acc.	Bias	Acc.	Bias	Acc.	Bias	Acc.	Bias
Prediction of Holstein or Jersey
Jersey	GBLUP	Holstein	0.09	0.55	0.10	0.54	0.09	0.63	0.15	0.42	0.17	0.48	0.12	0.51
Jersey	BayesR	Holstein	0.19	0.59	0.21	0.62	0.27	1.29	0.48	0.60	0.20	0.28	0.27	0.67
Holstein	GBLUP	Jersey	0.10	0.38	0.31	0.96	0.29	1.34	0.20	0.63	0.43	1.65	0.26	0.99
Holstein	BayesR	Jersey	0.09	0.21	0.30	0.53	0.26	0.71	0.33	0.56	0.42	0.79	0.28	0.56
Prediction of Australian Reds
Holstein	GBLUP	AustRed	0.10	0.42	0.10	0.27	−0.01	0.00	0.41	0.94	0.48	1.25	0.22	0.57
Holstein	BayesR	AustRed	0.20	0.67	0.19	0.53	0.04	0.17	0.52	0.92	0.44	0.79	0.28	0.61
Jersey	GBLUP	AustRed	0.14	1.01	0.01	0.07	0.11	0.88	0.20	0.61	0.19	0.49	0.13	0.61
Jersey	BayesR	AustRed	0.35	1.60	0.08	0.28	0.19	1.12	0.41	0.59	0.21	0.33	0.25	0.78
Hol/Jer	GBLUP	AustRed	0.17	0.75	0.11	0.32	0.04	0.16	0.46	1.06	0.48	1.17	0.25	0.69
Hol/Jer	BayesR	AustRed	0.26	0.89	0.22	0.56	0.10	0.38	0.53	0.88	0.43	0.67	0.30	0.67
Avg.¹	GBLUP		0.12	0.56	0.17	0.61	0.14	0.71	0.27	0.70	0.36	1.08	0.21	0.73
	BayesR		0.18	0.56	0.25	0.57	0.21	0.79	0.44	0.68	0.35	0.58	0.28	0.64

FY = fat yield (kg/lactation), MY = milk yield (L/lactation), PY = protein yield (kg/lactation), F% = fat percentage (%) and P% = protein percentage in milk (%); Acc. = accuracy, measured as r(ŷ, y), where ŷ is the prediction of genetic merit. Bias = bias of the prediction, measured as the regression coefficient, b (ŷ, y); standard errors are approximately \( \frac{1}{\sqrt{262}}=0.062 \) for the Holstein predictions, \( \frac{1}{\sqrt{105}}=0.098 \) for Jersey predictions, \( \frac{1}{\sqrt{180}}=0.074 \) for Australian Red predictions (average of the predictions for cow and bull validation sets; accuracies for each Australian Red bull and cow sets are in Additional file 3: Table S4 (see Additional file 3: Table S4); ¹average across-breed prediction accuracy for GBLUP and BayesR is calculated using the average of the Australian Red predictions from the multi-breed Holstein/Jersey reference population, Jersey predictions from the Holstein reference population and Holstein predictions from the Jersey reference population.

ISSN: 1297-9686