The Efficiency of Genome-wide Selection for Genetic Improvement of Net Merit.

Abstract

Four methods of selection for net merit comprising two correlated traits were compared in this study: 1) EBV-only index (I(1)) which consists of the estimated breeding value (EBV) of both traits (i.e., traditional two-trait BLUP selection); 2) GEBV-only index (I(2)) which comprises the genomic estimated breeding values (GEBV) of both traits; 3) GEBV-assisted index (I(3)) which combines both the EBV and the GEBV of both traits; and 4) GBV-assisted index (I(4)) which combines both the EBV and the true genomic breeding value (GBV) of both traits. Comparison of these indices were based on three evaluation criteria (selection accuracy, genetic response (ΔH), and relative efficiency) under 64 scenarios that arise from combining two levels of genetic correlation (r(G)), two ratios of genetic variances between traits, two ratios of the genomic variance to total genetic variances for trait 1, four accuracies of EBV, and two proportions of r(G) explained by the GBV. Both selection accuracy and genetic responses of the indices I(1), I(3) and I(4) increased as the accuracy of EBV increased, but the efficiency of the indices I(3) and I(4) relative to I(1) decreased as the accuracy of EBV increased. The relative efficiency of both I(3) and I(4) was generally greater when the accuracy of EBV was 0.6 than when it was 0.9, suggesting that the genomic markers are most useful to assist selection when the accuracy of EBV is low. The GBV-assisted index I(4) was superior to the GEBV-assisted I(3) in all 64 cases examined, indicating the importance of improving the accuracy of prediction of genomic breeding values. Other parameters being identical, increasing the genetic variance of a high heritability trait would increase the genetic response of the genomic indices (I(2), I(3) and I(4)). The genetic responses to I(2), I(3) and I(4) was greater when the genetic correlation between traits was positive (r(G) =0.5) than when it was negative (r(G) =-0.5). The results of this study indicate that the effectiveness of the GEBV-assisted index I(3) is affected by heritability of and genetic correlation between traits, the ratio of genetic variances between traits, the genomic-genetic variance ratio of each index trait, the proportion of genetic correlation accounted by the genomic markers, and the accuracy of predictions of both EBV and GBV. However, most of these affecting factors are genetic characteristics of a population which are beyond the control of the breeders. The key factor subject to manipulation is to maximize both the proportion of the genetic variance explained by GEBV and the accuracy of both GEBV and EBV. The developed procedures provide means to investigate the efficiency of various genomic indices for any given combination of the genetic factors studied.