Genomic models for accurate estimation of breeding values in Sitka spruce (Picea sitchensis Bong. Carr)

Abstract Genomics is advancing estimation of breeding values for selective breeding. Traditional pedigree information used in the estimation of breeding values can still be competitive to genomic information when markers do not adequately capture genetic variance or genomic datasets are not sufficiently large. Using the UK Sitka spruce breeding population as a study system, we assessed the accuracy of breeding values using pedigree or genomic information. Nine phenotypic traits associated with growth and wood quality were measured across 1295 individuals in 50 full-sib families replicated across two plantation sites. Pedigree-based estimation was compared to genome-based estimation using identity by state (IBS, across families) or identity by descent (IBD, within families) information from 4123 SNPs. Accuracy of breeding values was assessed using prediction standard errors obtained from the fitted model, which assumes the model and parameters are correct, and a less biased 12-fold cross-validation by withholding phenotypes from the model fitting. Accuracies obtained directly from the model were very similar across different types of information, but the more objective cross-validation assessment showed genomic information gave the best predictions for all but one of the traits. No bias was observed for any of the trait- information combinations. Results demonstrate that the predictive performance of genomic information is already competitive with the pedigree information, despite evidence that the marker panel did not capture the full genetic variance. Furthermore, accuracies with genomic information have the potential to significantly improve, unlike pedigree, by increases in the size of the training set and marker density.