Genomic prediction (GP) or genomic selection is a method to predict the accumulative effect of all quantitative trait loci (QTLs) in a population by estimating the realized genomic relationships between the individuals and by capturing the linkage disequilibrium between markers and QTLs. Thus, marker preselection is considered a promising method to capture Mendelian segregation effects. Using QTLs detected in a genome-wide association study (GWAS) may improve GP. Here, we performed GWAS and GP in a population with 904 clones from 32 full-sib families using a newly developed 50 k SNP Norway spruce array. Through GWAS we identified 41 SNPs associated with budburst stage (BB) and the largest effect association explained 5.1% of the phenotypic variation (PVE). For the other five traits such as growth and wood quality traits, only 2 – 13 associations were observed and the PVE of the strongest effects ranged from 1.2% to 2.0%. GP using approximately 100 preselected SNPs, based on the smallest p-values from GWAS showed the greatest predictive ability (PA) for the trait BB. For the other traits, a preselection of 2000–4000 SNPs, was found to offer the best model fit according to the Akaike information criterion being minimized. But PA-magnitudes from GP using such selections were still similar to that of GP using all markers. Analyses on both real-life and simulated data also showed that the inclusion of a large QTL SNP in the model as a fixed effect could improve PA and accuracy of GP provided that the PVE of the QTL was ≥ 2.5%.

Genomic prediction (GP) or genomic selection is a method to predict the accumulative effect of all quantitative trait loci (QTLs) effects by capturing the linkage disequilibrium between markers and QTLs. Thus, marker preselection is considered a promising method to capture Mendelian segregation effects, especially for an oligogenic trait. Using QTLs detected in the genome-wide association study (GWAS) could improve genomic prediction, including informative marker selection and adding a QTL with the largest effect size as a fixed effect. Here, we performed GWAS and genomic selection studies in a population with 904 clones from 32 full-sib families using a newly developed 50k SNP Norway spruce array. In total, GWAS identified 41 SNPs associated with budburst stage (BB) and the SNP with the largest effect size explained 5.1% of the phenotypic variation (PVE). For the other five traits like growth and wood quality traits, only 2–13 SNPs were detected and PVE of the strongest effects ranged from 1.2–2.0%. GP with approximately 100 preselected SNPs based on the smallest p-values from GWAS showed the largest predictive ability (PA) for the oligogenic trait BB. But for the other polygenic traits, approximate 2000–4000 preselected SNPs, indicated by the smallest Akaike information criterion to offer the best model fit, still resulted in PA being similar to that of GP models using all markers. Analyses on both real-life and simulated data also showed that the inclusion of a large QTL SNP in the model as a fixed effect could improve PA and accuracy of GP provided that the PVE of the QTL was ≥ 2.5%.

Genome-wide association (GWA) has been used to detect quantitative trait loci (QTLs) in plant and animal species. Genomic prediction is to predict an accumulative effect of all QTL effects by capturing the linkage disequilibrium between markers and QTLs. Thus, marker preselection is considered a promising method to capture Mendelian segregation effects, especially for an oligogenic trait. Using QTLs detected in the GWA study could improve genomic prediction (GP), including informative marker selection and adding a QTL with the largest effect size as a fixed effect. In this study, we performed GWA and GP studies in a population with 904 clones from 32 full-sib families planted in four trials using a newly developed 50k SNP Norway spruce array. In total, GWAS identified 41 SNPs associated with the budburst stage (BB) and the SNP with the largest effect size explained 5.1% of the phenotypic variation (PVE). For other traits like height at tree ages six and 12, diameter at breast height, frost damage and Pilodyn penetration, only 2 - 13 SNP associations were detected and the PVE of the strongest effects ranged from 1.2% to 2.0%. GP with approximately 100 preselected SNPs based on the smallest p-values from GWAS showed the largest predictive ability (PA) for the oligogenic trait BB. But for the other polygenic traits, approximate 2000-4000 preselected SNPs, indicated by the smallest Akaike information criterion (AIC) to offer the best model fit, still resulted in PA being similar to that of GP models using all markers. Analyses on both real-life and simulated data also showed that the inclusion of a large QTL SNP in the model as a fixed effect could improve the PA and accuracy of GP provided that the PVE of the QTL was >=2.5%.