Genomic prediction of breeding values using previously estimated SNP variances

Table 2 Reliabilities of GEBV ¹

Trait	Scenario	BSSVS	BayesC	RR-BLUP	BLUP-SSVS	BLUP-C
Protein	FULL	0.480	0.464	0.409	0.468	0.458
	RAN50	0.294	0.274	0.257	0.477	0.469
	TOP50	0.345	0.336	0.307	0.473	0.459
	BOT50	0.119	0.121	0.106	0.479	0.467
UD	FULL	0.510	0.511	0.471	0.502	0.509
	RAN50	0.374	0.374	0.363	0.507	0.511
	TOP50	0.325	0.326	0.366	0.494	0.496
	BOT50	0.085	0.073	0.091	0.493	0.491
SCS	FULL	0.572	0.581	0.544	0.573	0.577
	RAN50	0.412	0.410	0.394	0.572	0.571
	TOP50	0.434	0.432	0.449	0.563	0.562
	BOT50	0.086	0.089	0.138	0.561	0.562
IFL	FULL	0.534	0.534	0.470	0.527	0.532
	RAN50	0.432	0.434	0.399	0.530	0.532
	TOP50	0.110	0.114	0.114	0.520	0.521
	BOT50	0.331	0.329	0.256	0.521	0.522
DLO	FULL	0.396^a	0.397^a	0.309^b	0.389^a,b	0.388^a,b
	RAN50	0.205	0.213	0.173	0.392^a,b	0.394^a,b
	TOP50	0.330	0.331	0.289	0.411^a	0.412^a
	BOT50	0.018	0.022	0.028	0.407^a	0.409^a
LON	FULL	0.417^a,b	0.419^a,b	0.341^a	0.409^a,b	0.409^a,b
	RAN50	0.282	0.280	0.227	0.418^a,b	0.422^a,b
	TOP50	0.354	0.353	0.320	0.434^b	0.436^b
	BOT50	0.014	0.023	0.032	0.428^a,b	0.429^a,b

Reliabilities are computed for six traits, five different models and four training scenarios using all (FULL), at random 50% (RAN50), the best 50% (TOP50), or the worst 50% (BOT50) of the training dataset.
¹Standard errors of reliabilities were on average equal to 0.029 and ranged from 0.010 to 0.034; ^a,bvalues with different superscripts indicate significant differences at P < 0.05; reliabilities of BSSVS, BayesC and RR-BLUP were compared to each other within the same scenario; reliabilities of BLUP-SSVS and BLUP-C for all four scenarios were always compared to reliabilities of BSSVS, BayesC and RR-BLUP obtained in the FULL scenario, because BLUP-SSVS and BLUP-C always used all training animals.

ISSN: 1297-9686