Accuracy of Weighted Single-Step Genomic Prediction for Milk Production Traits in Korean Holstein Cattle

Milk production traits are economically important indicators of dairy cattle performance, and genetic improvement depends on accurate breeding value prediction. This study compared pedigree-based BLUP (PBLUP), genomic BLUP (GBLUP), single-step genomic BLUP (ssGBLUP), and weighted single-step genomic BLUP (wssGBLUP) for 305-day milk yield (Milk305), fat yield (Fat305), and protein yield (Prot305) in Korean Holstein cattle. A total of 2267 genotyped animals were analyzed using a 50K SNP panel, with missing genotypes imputed using BEAGLE v5.5. Accuracy was calculated from the prediction error variance (PEV) of estimated breeding values. Repeatability ranged from 0.41 to 0.45 across methods and traits. Overall, PEV-derived accuracy for Milk305 was 0.52 ± 0.14, 0.31 ± 0.08, 0.51 ± 0.13, and 0.51 ± 0.13 for PBLUP, GBLUP, ssGBLUP, and wssGBLUP, respectively. For Fat305, PEV-derived accuracy was 0.52 ± 0.14, 0.41 ± 0.10, 0.51 ± 0.13, and 0.51 ± 0.14. For Prot305, values were 0.50 ± 0.14, 0.33 ± 0.08, 0.27 ± 0.10, and 0.50 ± 0.13, respectively. These accuracy estimates were derived internally from the prediction error variance (PEV) of the mixed-model equations and have not been validated against independent data; they should therefore be interpreted as relative, model-based comparisons among methods rather than as externally validated predictive performance. Genomic-window analysis identified heterogeneous marker-explained additive genetic variance, with the BTA14 region near DGAT1 among the highest-variance regions for Milk305 and Fat305. These findings indicate that the benefits of single-step and weighted genomic evaluation depend on trait and animal group.