Weighted single-step genomic best linear unbiased predictor enhances the genomic prediction accuracy for milk citrate predicted by milk mid-infrared spectra of Holstein cows in early lactation

Abstract

Previous studies have shown that milk citrate predicted by milk mid-infrared (MIR) spectra is strongly affected by a few genomic regions. This study aimed to explore the effect of weighted single-step GBLUP on the accuracy of genomic prediction (GP) for MIR-predicted milk citrate in early-lactation Holstein cows. A total of 134,517 test-day predicted milk citrate collected within the first 50 DIM on 52,198 Holstein cows from the first 5 parities were used. There were 122,218 animals in the pedigree, of which 4,479 had genotypic data for 566,170 SNPs. Two datasets (partial and whole datasets) were used to verify whether the accuracy of GP is improved using the following different methods. The (genomic) estimated breeding values (EBV or GEBV) in the partial and whole datasets were estimated by pedigree-based BLUP (ABLUP), single-step GBLUP (ssGBLUP, pedigree-genomic combined using no weight for SNP), and weighted ssGBLUP (WssGBLUP, pedigree-genomic combined using weighted SNP), respectively. The difference between the 2 datasets is that the phenotypic data from 2017 to 2019 in the partial dataset were set as missing values. One hundred eighty-one youngest cows with genomic data were selected as the validation population. A linear regression method was used to compare EBV (GEBV) predicted for partial and whole datasets. The accuracies of GP for ABLUP and ssGBLUP were 0.42 and 0.70, respectively. The accuracies of GP for WssGBLUP in the 5 iterations with different CT (constant) values (determines departure from normality for SNP effects) ranged from 0.70 to 0.86. This study showed that weighted SNP is beneficial in improving prediction accuracy for predicted milk citrate.