Accuracy of genomic values predicted using deregressed predicted breeding values as response variables

Highly accurate predicted genetic values must be obtained at an early age to promote rapid genetic progress. The objectives of this study were to compare accuracies (R-2) of genomic values (GVs) and to estimate genetic correlation between true genetic values and genomic values obtained using predicted breeding values (EBV) and deregressed EBV (DEBV) as response variables. A first population, effective population size 800 and 100 generations, was simulated using the QMSim program to generate linkage disequilibrium. Thereafter, 20 males and 200 females were used to generate a second 14-generation population, with 6,400 individuals per generation and its corresponding phenotype and genotype in SNP terms. Generations 7 to 14 of the second population were used in several combinations as training (PEn) and evaluation (PEv) subpopulations. GVs, their accuracies, and genetic correlations were obtained using the GenSel and ASREML programs. When PEn was the largest, the mean R-2 of GV was the highest, 0.77 +/- 0.01. The closer PEn was to PEv, the higher the R-2, and correspondingly, the lower the predicted error variance. The trends for R-2 and PEV held true for both EBV and DEBV used as response variables. Genetic correlation estimates between true genetic values and GVs varied from 0.41 to 0.53 in the two scenarios studied. They decreased when PEn and PEv were farther apart. There were only slight advantages of using DEBVs as response variables over using EBVs.