Table 3.

Mean prediction accuracy comparison between conventional and multi-class Bayesian Alphabet for real pig data (single-trait) and real rice data (multiple-trait).

Data	Method	RR-BLUP	BayesA	BayesB	BayesCΠ	Bayesian LASSO	Ensemble
Pig	Conventional	0.516	0.565	0.568	0.532	0.517	0.550
	Multi-class	0.516	0.565	0.569	0.532	0.516	0.550
Rice	Conventional	0.378	0.353	0.372	0.384	0.378	0.377
	Multi-class	0.374	0.357	0.363	0.375	0.373	0.373

The comparison of mean prediction accuracy on the trait 13-week tenth rib backfat (mm) from pig data and trait FTA of rice real data across 50 validation datasets between multi-class Bayesian Alphabet using genome annotation information and conventional Bayesian Alphabet. The paired t-test (p < 0.1) was used to declare the significant difference. No significant differences were found between multi-class and conventional Bayesian Alphabet methods for both real pig and rice data (p < 0.1).