Table 2. Predictive ability (PA) of GBLUP and RKHS statistical models obtained with 28,091 SNPs selected based on linkage disequilibrium (LD), and based on association with target trait detected by genome wide association study (GWAS).

			Environment & size of the population
			280 accessions		204 accessions
			E1		E1		E2		E3
			LD	GWAS	LD	GWAS	LD	GWAS	LD	GWAS
GBLUP	DTF	PA	0.540 b	0.618 a	0.599 b	0.653 a	0.450 b	0.519 a	0.579 b	0.635 a
		STD	0.084	0.067	0.082	0.071	0.129	0.108	0.098	0.097
	GY	PA	0.389 b	0.489 a	0.345 b	0.398 b	0.420 b	0.476 b	0.672 b	0.682 a
		STD	0.09	0.061	0.115	0.124	0.12	0.162	0.122	0.133
	PH	PA	0.505 b	0.623 a	0.558 b	0.595 a	0.648 b	0.679 b	0.673 b	0.703 a
		STD	0.087	0.086	0.085	0.046	0.087	0.049	0.073	0.057
RKHS	DTF	PA	0.539 b	0.633 a	0.585 b	0.662 a	0.466 b	0.553 a	0.554 b	0.628 a
		STD	0.081	0.062	0.09	0.071	0.114	0.110	0.098	0.089
	GY	PA	0.387 b	0.490 a	0.352 b	0.391 b	0.435 b	0.479 b	0.668 b	0.695 b
		STD	0.089	0.054	0.106	0.127	0.113	0.683	0.149	0.132
	PH	PA	0.501 b	0.625 a	0.547 b	0.595a	0.646 b	0.683 a	0.675 b	0.711 a
		STD	0.09	0.088	0.084	0.046	0.078	0.051	0.075	0.060

DTF: days to flowering; GY: grain yield PH: plant height (PH); E1: lowland non-stress environment; E2: lowland stressed; E3: upland stressed. AP of GWAS-derived markers predictions that are followed by different letters than their LD-derived counterpart are significantly different at α <0.05.