Genomic Selection for growth traits in Eucalyptus: accuracy within and across breeding populations

Background Genomic selection (GS) involves selection decisions based on genomic breeding values estimated as the sum of the effects of genome-wide markers capturing most QTLs for the target trait(s). GS is revolutionizing breeding practice for complex trait in domestic animals. The same approach and concepts can be readily applied to forest tree breeding. Trees also have long generation times and late expressing traits. Differently from association genetics that aims at dissecting complex traits in their discrete components, GS precludes the discovery of individual marker-trait associations and focuses on prediction of performance. By capturing the “missing heritability” of complex quantitative traits beyond the few effect variants that association genetics has so far typically identified, GS might soon cause a paradigm shift in forest tree breeding. In a prior deterministic study we assessed the impact of linkage disequilibrium (modeled by Ne and inter-marker distance), the size of the training set, trait heritability and the number of QTL on the predicted accuracy of GS [1]. Results indicate that GS has the potential to radically improve the efficiency of tree breeding. The benchmark accuracy of conventional BLUP-based phenotypic selection (0.68) was reached by GS even at a marker density ~2 markers/cM when Ne ≤30, while up to 10 markers/cM are necessary for larger Ne. Shortening the breeding cycle by 50% with GS provides an expected increase ≥100% in selection efficiency. To validate these simulation results we carried out a large multi-population proof-of-concept study of GS in tropical Eucalyptus. In this report we present results of this on-going study for two populations and three different quantitative traits.