A comparison of five methods to predict genomic breeding values of dairy bulls from genome-wide SNP markers

Table 2 Summary of MBV prediction in the training data for five methods obtained by cross-validation

Trait	Method	MSEP		r_EBV,MBV		b_EBV,MBV
ASI	FR-LS	1,090	(124.4)	0.53	(0.036)	0.71	(0.048)
	RR-BLUP	712	(93.5)	0.71	(0.017)	1.07	(0.076)
	Bayes-R	714	(95.3)	0.71	(0.016)	1.09	(0.071)
	SVR	700	(92.2)	0.72	(0.017)	1.06	(0.079)
	PLSR	735	(95.4)	0.70	(0.022)	0.93	(0.069)
PPT	FR-LS	0.0135	(0.0023)	0.43	(0.089)	0.62	(0.155)
	RR-BLUP	0.0104	(0.0018)	0.56	(0.067)	1.01	(0.104)
	Bayes-R	0.0104	(0.0010)	0.56	(0.067)	1.06	(0.117)
	SVR	0.0100	(0.0010)	0.58	(0.064)	1.01	(0.100)
	PLSR	0.0109	(0.0012)	0.55	(0.061)	0.81	(0.078)

Mean square error (MSEP), correlation (r_EBV,MBV) between EBV and MBV, and regression coefficient (b_EBV,MBV) of EBV on MBV derived by 5-fold cross-validation of the training data set, standard errors in parentheses; ASI: Australian Selection Index; PPT: protein percentage; FR-LS: fixed regression-least squares; RR-BLUP: random regression-BLUP; Bayes-R: Bayesian regression; SVR: support vector regression; PLSR: partial least squares regression.

ISSN: 1297-9686