FIGURE 5.

Accuracies and bias of genomic predictions based on individual SNPs and haplotypes for the simulations of traits with moderate (A) and low (B) heritability (0.30 and 0.10, respectively). Breed_B, Breed_C, and Breed_E: simulated pure breeds with different genetic backgrounds; Comp_2 and Comp_3: composite breeds from two and three pure breeds, respectively. 600 K: high-density panel; 50 K: medium-density panel; IPS_LD01, IPS_LD03, and IPS_LD06: independent and pseudo-SNPs from blocks with LD thresholds of 0.1, 0.3, and 0.6, respectively, in a single genomic relationship matrix; PS_LD01, PS_LD03, and PS_LD06: only pseudo-SNPs from blocks with LD threshold of 0.1, 0.3, and 0.6, respectively; and IPS_2H_LD01, IPS_2H_LD03, and IPS_2H_LD06: independent and pseudo-SNPs from blocks with LD thresholds of 0.1, 0.3, and 0.6, respectively, in two genomic relationship matrices. Zero values for both accuracies and bias mean no results were obtained, due to poor quality of genomic information or no convergence of the genomic prediction models. The same lower-case letters mean no statistical difference comparing genomic prediction methods within population at 5% significance level based on the Tukey test.