Genomic selection on breeding time in a wild bird population

Abstract

Artificial selection experiments are a powerful tool in evolutionary biology. Selecting individuals based on multimarker genotypes (genomic selection) has several advantages over phenotype-based selection but has, so far, seen very limited use outside animal and plant breeding. Genomic selection depends on the markers tagging the causal loci that underlie the selected trait. Because the number of necessary markers depends, among other factors, on effective population size, genomic selection may be in practice not feasible in wild populations as most wild populations have much higher effective population sizes than domesticated populations. However, the current possibilities of cost-effective high-throughput genotyping could overcome this limitation and thereby make it possible to apply genomic selection also in wild populations. Using a unique dataset of about 2000 wild great tits (Parus major), a small passerine bird, genotyped on a 650 k SNP chip we calculated genomic breeding values for egg-laying date using the so-called GBLUP approach. In this approach, the pedigree-based relatedness matrix of an "animal model," a special form of the mixed model, is replaced by a marker-based relatedness matrix. Using the marker-based relatedness matrix, the model seemed better able to disentangle genetic and permanent environmental effects. We calculated the accuracy of genomic breeding values by correlating them to the phenotypes of individuals whose phenotypes were excluded from the analysis when estimating the genomic breeding values. The obtained accuracy was about 0.20, with very little effect of the used genomic relatedness estimator but a strong effect of the number of SNPs. The obtained accuracy is lower than typically seen in domesticated species but considerable for a trait with low heritability (∼0.2) as avian breeding time. Our results show that genomic selection is possible also in wild populations with potentially many applications, which we discuss here.

Keywords: Animal model; GBLUP; genomic breeding values; phenology; quantitative genetic.

Figure 1

Correspondence between pairwise relatedness estimates from the pedigree and SNPs. The genomic relatedness estimated using all SNPs and the vanRaden (2008) estimator (including pedigree information, see Methods for details) is plotted against the relatedness estimated from the pedigree information. The dotted line indicates the 1:1 relationship.

Figure 2

Estimated accuracies (and 95% confidence interval) for EBVs and GEBVs of egg‐laying date for four different estimation methods (ped: pedigree only, vR: vanRaden (2008), Y: Yang et al. (2010), vR‐ped: vanRaden (2008) including the pedigree in the estimation, Y‐ped: Yang et al. (2010) including the pedigree in the estimation). See Methods for details on estimation procedure.

Figure 3

Accuracy of GEBVs (and 95% confidence interval) as a function of the number of markers used to estimate genomic relatedness (using VanRaden 2008).