Genomic Prediction of Meat Tenderness in Nellore Cattle: Multi‐Trait and Weighted Single‐Step Genomic Best Linear Unbiased Prediction Approaches

This study aimed to evaluate the impact of different genomic prediction approaches on the predictive ability for meat tenderness in Nellore cattle. Phenotypic (n = 73,286), pedigree (n = 4,141,892) and genomic (n = 15,300) data from animals belonging to the genetic improvement program of the National Association of Breeders and Researchers (ANCP) were used. Six models were tested: (1) standard ssGBLUP (Single-step Genomic Best Linear Unbiased Prediction), considering direct additive genetic and residual effects as random, contemporary group (CG) as a fixed effect, and slaughter age as a linear and quadratic covariate; (2) Model 1 + ssGBLUP weighted with SNP effects from the first WssGWAS iteration; (3) Model 1 + ssGBLUP weighted with SNP effects from the second WssGWAS iteration; (4) Model 1 + body weight as a covariate; (5) Model 1 as a bi-trait model with body weight at 450 days (W450); (6) Model 1 as a multi-trait model with carcass traits: ribeye area (REA), backfat thickness (BFT) and rump fat thickness (RFT). Predictive ability was evaluated using linear regression, in which the dataset was divided into a complete and a partial subset (n = 374) dataset. Accuracy ranged from 0.04 (Models 2 and 3) to 0.37 (Model 6). Bias was low for all models, with Models 2 and 3 showing the least bias (-0.001). Model 6 showed the best performance in terms of accuracy and correlation (0.897), suggesting it was more effective in capturing genetic variability of meat tenderness, while reducing bias and increasing the precision of the estimates. Multi-trait models may offer more robust genomic predictions by leveraging trait correlations to increase prediction accuracy.