Inclusion of Dominance and GxE Effects in Genomic Selection Model to Improves Predictive Ability

Dominance effects are often neglected in genome selection model, adding to the residual effect. Another factor normaly omited from these models are genotype by environment (GxE) interactions becuase their estimation  requires collecting phenotypic data from the training population in multiple environments. The inclusion of these effects in the model may improve their predictive ability, particularly for forestry crops like loblolly pine that are established accross a highly heterogeneous landscape. The aim of current study was to assess the predictive ability of height in loblolly pine using a model for genomic selection including dominance and GxE interaction effects. The analysis used GBLUP in the loblolly pine CCLONES (Comparing Clonal Lines On Experimental Sites) population, which is composed of 923 individuals genotyped with 4,722 polymorphic SNPs. CCLONES is established in four sites along a north-south transect: Palatka and Nassau (Florida,USA), Cuthbert and the B.F. Grant Forest (Georgia, USA). The predictive ability was estimated by the average of 10-fold of correlations between genomic breeding values and phenotypes. Compared to the additive model there was a modest increase in the predictive ability when using the additive-dominance model, when both of the models include GxE interaction effect.