Multi-trait Improvement by Predicting Genetic Correlations in Breeding Crosses

The many quantitative traits of interest to plant breeders are often genetically correlated, which can complicate progress from selection. Improving multiple traits may be enhanced by identifying parent combinations – an important breeding step – that will deliver more favorable genetic correlations (r(G)). Modeling the segregation of genomewide markers with estimated effects may be one method of predicting r(G) in a cross, but this approach remains untested. Our objectives were to: (i) use simulations to assess the accuracy of genomewide predictions of r(G) and the long-term response to selection when selecting crosses on the basis of such predictions; and (ii) empirically measure the ability to predict genetic correlations using data from a barley (Hordeum vulgare L.) breeding program. Using simulations, we found that the accuracy to predict r(G) was generally moderate and influenced by trait heritability, population size, and genetic correlation architecture (i.e., pleiotropy or linkage disequilibrium). Among 26 barley breeding populations, the empirical prediction accuracy of r(G) was low (-0.012) to moderate (0.42), depending on trait complexity. Within a simulated plant breeding program employing indirect selection, choosing crosses based on predicted r(G) increased multi-trait genetic gain by 11–27% compared to selection on the predicted cross mean. Importantly, when the starting genetic correlation was negative, such cross selection mitigated or prevented an unfavorable response in the trait under indirect selection. Prioritizing crosses based on predicted genetic correlation can be a feasible and effective method of improving unfavorably correlated traits in breeding programs.