Estimation & Prediction

Variation in genomic prediction accuracy according to trait definition and validation strategy in Nellore cattle

---

Abstract

This study evaluated the performance of different genomic prediction models for the productive efficiency trait in Nellore beef cattle. Records were available for the calf weaning weight to cow weight ratio, defined as a calf trait (WR, n = 64,505) and as a cow trait (WRC, n = 34,481). The pedigree comprised 492,042 animals (with 9,632 sires and 165,757 dams), including 25,000 genotyped individuals with 93,400 SNPs. For validation purposes, two partial datasets were created: (1) exclusion of recent phenotypes (young-animal validation: calves born in 2021-2022 for WR and cows born in 2019-2020 for WRC), and (2) exclusion of phenotypes from genotyped sires (with 15-200 offspring) and their progeny (progeny-test validation). Validation populations included 1,440 young animals and 163 sires for WR, and 594 cows and 40 sires for WRC. Genomic and pedigree-based estimated breeding values (EBV or GEBV) in the partial and complete datasets were obtained using pedigree-based best linear unbiased prediction (PBLUP), single-step genomic BLUP (ssGBLUP), and weighted ssGBLUP (WssGBLUP), fitted under single-trait models with maternal (WR) and permanent environmental effects (WRC). Predictive ability was assessed by linear regression of EBV (GEBV) from partial and complete datasets, using bias, dispersion, and accuracy as criteria. For WR, under young-animal validation, including genomic data reduced bias from -0.21 (PBLUP) to -0.07 (single-step models), brought dispersion closer to unity, and increased accuracy by up to 0.18 points (66.7%). In progeny-test validation, bias was similar among methods (0.05-0.07), but accuracies increased by up to 0.16 points (64%) with genomic models. For WRC, young-animals validation showed similar bias across models (-0.53 to -0.55), and an accuracy increase of about 0.13 points (50%) with genomic models, while dispersion remained low. In progeny-test validation, bias decreased from 0.85 (PBLUP) to 0.71 (ssGBLUP), but accuracy gains were negligible (0.28 with pedigree to - 0.30 with single-step models). In conclusion, the benefits of incorporating genomic information varied depending on the trait and validation strategy. The performance of ssGBLUP and WssGBLUP was similar across analyses. Greater improvements were observed for WR under young-animal validation, whereas gains for WRC were modest. These results highlight that the impact of genomics depends on the biological nature of the trait and the structure of the validation data.

Keywords: 2026

How to Cite:

Magaço, F., Miranda, C., Pereira, D., Fragomeni, B., Menezes, G., Carvalho Filho, I., Toral, F. & Pereira, I., (2026) “Variation in genomic prediction accuracy according to trait definition and validation strategy in Nellore cattle”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286580. doi: https://doi.org/10.31274/wcgalp.24055

Rights:

1