Fig. 2
Predictive ability obtained with GBLUP using different weighting factors based on high-density array data and whole-genome sequencing data. The predictive ability was measured as the correlation between direct genomic breeding values (DGV) and de-regressed proofs (DRP) in the validation set. Results are for the traits eggshell strength (ES), feed intake (FI) and arcsine transformed laying rate in the last third of the laying period (LR). HD stands for high-density data and WGS stands for whole-genome sequencing data. G I stands for the original GBLUP, G P represents the model with −(log10 P) from GWAS as weighting factors where principal components that were significant at P values ≤ 10−100 (≤ 0.05) were used as covariates in the GWAS model, denoted as P 100 (P 005). G S stands for the GBLUP model with the squares of estimated SNPs effect as weighing factors. G z stands for the results from BLUP|GA as described in the “Methods” section. G G stands for the original GBLUP but only based on genic SNPs. The dashed horizontal line denotes the median predictive ability of GBLUP with HD data as a reference. Note that all the outliers for trait LR were from the same replicate