Balanced selection on purebred and crossbred performance increases gain in crossbreds

Abstract

Background: Genomic selection can be applied to select purebreds for crossbred performance (CP). The average performance of crossbreds can be considered as the summation of two components, i.e. the breed average (BA) of the parental breeds and heterosis (H) present in crossbreds. Selection of pure breeds for CP based on genomic estimated breeding values for crossbred performance (GEBV-C) or for purebred performance (GEBV-P) may differ in their ability to exploit BA and H and can affect the merit of crossbreds in both the short and long term. Selection based on GEBV-C is beneficial for CP, because H in crossbreds is efficiently exploited, whereas selection on GEBV-P results in more genetic progress in pure breeds, which increases the BA component of CP. To investigate the outcome of selection on GEBV-C and GEBV-P in both the short and long term, a two-way crossbreeding program was simulated to test the following hypotheses: (1) does selection on GEBV-P result in higher long-term CP compared to selection on GEBV-C and (2) does selection on a combination of GEBV-P and GEBV-C lead to more long-term gain in CP than selection on either separately.

Methods: We investigated the performance of crossbreds in a two-way crossbreeding program across 40 generations and considered different criteria to select purebred parents that ranged from selection on purebred performance to selection for CP with different weights on genomic evaluations based on purebred and CP. These criteria were compared under three genetic models to investigate the effects of the amount of dominance variance, absence of over-dominance, and the structure of the reference population on CP, both in the short and long term.

Results and conclusions: Although beneficial in the short to medium term, genomic selection in pure breeds on a criterion that specifically targets CP was inferior to selection for purebred performance in the long term. A selection criterion that maximizes a combination of short- and long-term responses in CP, should improve the components that define crossbred merit (i.e., BA and H) simultaneously.

Fig. 1

Mean phenotype of crossbred animals for different selection criteria relative to the reference selection criteria (w = 0). Mean phenotype of crossbreds for each selection criterion was plotted relative to the reference selection criterion (w = 0). The general criterion for selection of purebred parents was $S{C}_i=\left(1-w\right)·{\text{GEBV}}_{\mathit{iP}}+w·{\text{GEBV}}_{\mathit{iC}}$. Training in Models 2 and 3 was on purebred animals. OD = % QTL with over-dominance

Fig. 2

Cumulative response to selection. The general criterion for selection of purebred parents was $S{C}_i=\left(1-w\right)·{\text{GEBV}}_{\mathit{iP}}+w·{\text{GEBV}}_{\mathit{iC}}$. Training under genetic Models 2 and 3 was on purebred animals. OD =  % QTL with over-dominance. Inset plots show the mean performance of crossbred animals under each model for generations 30 to 40

Fig. 3

Mean phenotype of crossbred animals over generations under genetic Model 1. Two extreme values of w in the selection criteria (w = 0 and 1) were compared for purebred and crossbred training

Fig. 4

Mean phenotypic average of pure breeds. Mean phenotype of pure breeds are expressed relative to the means obtained with the reference selection criterion (w = 0). The general criterion for selection of purebred parents was $S{C}_i=\left(1-w\right)·{\text{GEBV}}_{\mathit{iP}}+w·{\text{GEBV}}_{\mathit{iC}}$. Training under genetic Models 2 and 3 was on purebred data. OD = % QTL with over-dominance

Fig. 5

Heterosis in crossbreds. The amount of heterosis obtained with each selection criterion was calculated as the difference between crossbred performance (CP) and breed average (BA) over generations. The general criterion for selection of purebred parents was $S{C}_i=\left(1-w\right)·{\text{GEBV}}_{\mathit{iP}}+w·{\text{GEBV}}_{\mathit{iC}}$. Training under genetic Models 2 and 3 was on purebred data. OD = % QTL with over-dominance

Fig. 6

Fixation of over-dominant QTL in the parental breeds. The percentages (inside bars) indicate the proportion of fixed over-dominant QTL that are fixed for the same or alternate alleles in the parental breeds. Bolded blue values above bars indicate the total fixation of over-dominant QTL in generation 40 for each selection criteria. The general criterions for selection of purebred parents was $S{C}_i=\left(1-w\right)·{\text{GEBV}}_{\mathit{iP}}+w·{\text{GEBV}}_{\mathit{iC}}$

Fig. 7

Accuracy of selection for five selection criteria for each genetic model. Accuracies are presented for breed A only. M1 (PT): genetic Model 1 with purebred training, M1 (CT): genetic Model 1 with crossbred training

Fig. 8

Correlation of LD phase (R_XY) between populations for SNP pairs for different distances between SNPs