Genetic Evaluation of a Loblolly Pine Population Using Hybrid Genetic Relationship Matrices

A cloned population of Pinus taeda and simulated data were used to compare accuracies of genomic estimated breeding values based on the blended genomic and pedigree relationship matrices. We used 3461 SNP markers to construct realized genomic relationships using different allele frequency methods.  Genomic relationship matrices (G) based on SNP markers or simulated data produced similar high accuracy values of predictions for the genotyped individuals. However, compared to the traditional genetic evaluations based on pedigree and phenotype, we observed higher accuracies when hybrid genetic relationships matrices were used in a BLUP approach for simulated data but not for empirical data. The hybrid genetic relationships matrices provided segregation of Mendelian sampling effect within full-sib crosses when phenotypes were not available in validation populations. Regressing the true breeding values on the genomic estimated breeding values produced little bias for the simulation data set.  For the empirical marker data, we observed significant differences in bias between G matrices. Genomic relationship based on observed allele frequencies produced the least bias. Scaling of the G matrix based on observed allele frequencies did not have an effect on the accuracy of predictions. Blending genetic relationships obtained from pedigree and DNA markers is an attractive strategy for forest tree breeding because breeders can capture Mendelian segregation effect, increase accuracy of predictions and also account for experimental design in a classical BLUP approach.