. 2021 Mar 4;12:645111. doi: 10.3389/fpls.2021.645111

TABLE 3.

Across population prediction accuracy of GS models.

Training population^a	Test population	Trait	Model^b
			GBLUP-A	GBLUP-AD	BB-A	BB-AD	BL-A	BL-AD	RKHS	RF
105-ILs	275-F₁s	Petiole length	0.352	0.364	0.326	0.338	0.369	0.375	0.359	0.343
		Leaf area	0.440	0.466	0.383	0.415	0.461	0.481	0.442	0.463
		Brix	0.167	0.214	0.087	0.148	0.268	0.284	0.241	0.254
		Fruit hardness	0.557	0.570	0.588	0.591	0.454	0.455	0.582	0.487
		Pericarp color	0.430	0.413	0.438	0.425	0.412	0.403	0.435	0.400
275-F₁s	105-ILs	Petiole length	0.309	0.335	0.330	0.350	0.297	0.336	0.338	0.355
		Leaf area	0.282	0.333	0.296	0.364	0.261	0.318	0.317	0.402
		Brix	−0.049	−0.024	−0.061	−0.036	−0.043	0.013	−0.029	0.021
		Fruit hardness	0.497	0.517	0.522	0.543	0.482	0.501	0.499	0.467
		Pericarp color	0.302	0.310	0.341	0.344	0.279	0.287	0.299	0.316

The accuracy was evaluated as a Pearson’s correlation coefficient between phenotypic and predicted values.^a105-ILs, 105 inbred lines; 275-F₁s, 275 test F₁ hybrids.^bGBLUP, Genomic best linear unbiased prediction; BB, Bayes B; BL, Bayesian Lasso; RKHS, Reproducing kernel Hilbert space regression; RF, Random forest.; -A, additive effect model; -AD, additive plus dominant effect model.