Efficiency of augmented and double-replicated designs under an array of genetical and spatial scenarios

Abstract

Early stage breeding trials evaluate hundreds or thousands of genotypes in one or more sites, requiring experiments which often present genotypes with low or no replication. Some of the proposed designs are augmented design (AD), and double replicated (DR, where all genotypes have two replications). In this work, computer simulations were used to evaluate the performance of designs with no or limited replication to be used for estimation of variance components, heritability, genotypic and breeding values (BV), and their achieved genetic gains under an array of genetical and spatial scenarios. Results from all simulation evaluated indicate that spatial analysis provided superior results than a no-spatial model. The incorporation of nugget effect brought better accuracies, even when there is no nugget on the field. Incorporation of pedigree information on the estimation of BV resulted on accuracy gains with good correlations between true and predicted BVs. However, the benefits of spatial analyses were less relevant once pedigree was incorporated. DR designs presented better performance than AD; and AD designs with 6.25% control plots showed similar results than AD with 25% control plots. In summary, the increase of only one replication, conferring two plots per genotype provided with very good estimations of genetic values. However, unreplicated trials provides with reasonable estimates when under spatial analyses and/or when they incorporate pedigree information.