Impact of prior specifications in a shrinkage-inducing Bayesian model for quantitative trait mapping and genomic prediction

Table 4 Comparison of genomic estimated breeding values (GEBV) and true breeding values (TBV) and predictive ability via cross-validation under varying prior specifications for the QTLMAS XII data

Prior specification		Accuracya		Rank correlationb		Biasc		GEM Cross-validationd
p ₀	b	MCMC	GEM	MCMC	GEM	MCMC	GEM	r _I	r _II	b _I	b _II
0.99	0.01	0.87	0.80	0.56	0.42	0.91	0.87	0.92	0.93	0.97	0.98
0.99	0.001	0.88	0.87	0.49	0.51	0.88	0.86	0.94	0.94	0.97	0.97
0.999	0.01	0.88	0.79	0.57	0.41	0.94	0.88	0.91	0.93	0.99	1.00
0.999	0.001	0.90	0.84	0.54	0.51	0.92	0.84	0.92	0.91	0.96	0.95
Combined ^e		0.89	0.88	0.56	0.53	0.94	0.98	0.96	0.96	1.05	1.04

^aPearson correlation between GEBV and TBV, for MCMC chains and generalized EM algorithm (GEM);
^bSpearman rank correlation between GEBV and TBV for the 10% of the individuals with highest TBV;
^cSlope Coefficient from regressing TBV on GEBV;
^dAccuracy (r: Pearson correlation of GEBV and phenotype, divided by the square root of heritability) and bias (b: slope coefficient from regressing the phenotype on GEBV) as estimated from two 10-fold cross-validation approaches: (I) entire families assigned together to folds and (II) individuals from the same family assigned to different folds;
^eCombined estimates obtained by averaging GEBVs across prior specifications.

ISSN: 1297-9686