Factors affecting the accuracy of genomic selection for growth and wood quality traits in an advanced-breeding population of black spruce (Picea mariana)

Abstract

BackgroundGenomic selection (GS) uses information from genomic signatures consisting of thousands of genetic markers to predict complex traits. As such, GS represents a promising approach to accelerate tree breeding, which is especially relevant for the genetic improvement of boreal conifers characterized by long breeding cycles. In the present study, we tested GS in an advanced-breeding population of the boreal black spruce (Picea mariana [Mill.] BSP) for growth and wood quality traits, and concurrently examined factors affecting GS model accuracy.ResultsThe study relied on 734 25-year-old trees belonging to 34 full-sib families derived from 27 parents and that were established on two contrasting sites. Genomic profiles were obtained from 4993 Single Nucleotide Polymorphisms (SNPs) representative of as many gene loci distributed among the 12 linkage groups common to spruce. GS models were obtained for four growth and wood traits. Validation using independent sets of trees showed that GS model accuracy was high, related to trait heritability and equivalent to that of conventional pedigree-based models. In forward selection, gains per unit of time were three times higher with the GS approach than with conventional selection. In addition, models were also accurate across sites, indicating little genotype-by-environment interaction in the area investigated. Using information from half-sibs instead of full-sibs led to a significant reduction in model accuracy, indicating that the inclusion of relatedness in the model contributed to its higher accuracies. About 500 to 1000 markers were sufficient to obtain GS model accuracy almost equivalent to that obtained with all markers, whether they were well spread across the genome or from a single linkage group, further confirming the implication of relatedness and potential long-range linkage disequilibrium (LD) in the high accuracy estimates obtained. Only slightly higher model accuracy was obtained when using marker subsets that were identified to carry large effects, indicating a minor role for short-range LD in this population.ConclusionsThis study supports the integration of GS models in advanced-generation tree breeding programs, given that high genomic prediction accuracy was obtained with a relatively small number of markers due to high relatedness and family structure in the population. In boreal spruce breeding programs and similar ones with long breeding cycles, much larger gain per unit of time can be obtained from genomic selection at an early age than by the conventional approach. GS thus appears highly profitable, especially in the context of forward selection in species which are amenable to mass vegetative propagation of selected stock, such as spruces.

Accuracy of genomic selection for growth and wood quality traits in two control-pollinated progeny trials using exome capture as genotyping platform in Norway spruce

“…We found that the accuracy improved from the TS/VS ratios of 1:1 to 3:1, but accuracy only improved a little after the TS/VS ratio of 3:1. This is different from other studies [4,14] that increased ratio of the TS/VS beyond 3:1 still increase the accuracy. However, we found that our result concurs with other reports when the ratio of TS/VS is related to the number of trees per family [14].…”

“…5). Similar results were obtained by other studies [11,14,41,49]. However, the relative efficiency between ABLUP and GBLUP (the accuracy ratio) is more or less similar in both full-sib and half-sib populations, which indicates that GS could be used in both half-sib and full-sib populations.…”

“…We found that our genomic accuracy for tree height and Pilodyn using half-sib family was lower than that reported by Lenz et al [14] in black spruce, even though more SNPs were used in this study (116K in 20,695 contigs vs 0.49K from SNP chip). The difference is likely due to lower heritabilities (i.e.…”

“…Accuracy and PA obtained from the subset of markers with the largest positive effects were slightly higher than those from a subset of markers based on random selection, implying that using the subset of markers with the largest effects in genomic regions with small LD decay could track relationship more effectively than random markers. It also implies that using the subset of markers with the largest positive effect could also obtain some effects based on part of the short-range LD [14,41]. Thus, this factor could also be potentially useful to reduce genotyping the number of SNPs and GS cost, in highly structured population when the genome location of markers is known.…”

CVRMS: Cross-validated Rank-based Marker Selection for Genome-wide Prediction of Low Heritability

“…The best subset of features contains the least number of dimensions that most contribute to prediction accuracy(Guyon et al, 2003). Several studies(Bermingham et al, 2015, Li et al, 2018, and Lenz et al, 2017 have performed marker selection by selecting most significant SNPs through the result of a genome-wide association study (GWAS) or marker effect analysis, but selecting a marker in this way only considers the relationship between the marker and the trait, not the correlation among the markers.…”