Genomic Prediction Ability Using Linear and Threshold Approaches for Different Stayability Definitions in Nellore Females

Abstract

Enhancing female longevity through regular calvings improves herd replacement rates and reduces the costs associated with replacing low reproductive efficiency animals. Stayability (STAY), defined as a cow's ability to remain productive in the herd, is a categorical trait that challenges traditional genetic evaluation due to its non-normal distribution. This study aimed to estimate genomic predictions for different STAY definitions—based on the number of calvings at specific ages in Nellore females—and to compare the predictive ability of linear and threshold models using the linear regression (LR) method. Phenotypic and genotypic data from 187 herds provided by the Nellore Brazil breeding program (ANCP, Ribeirão Preto, Brazil) were used. Four STAY definitions (STAY48-2, STAY48-3, STAY54-2, STAY54-3, STAY72-3) were evaluated. Genomic estimated breeding values (GEBV) were obtained using univariate linear and threshold models implemented in the BLUPF90 software family. Variance components were transformed from liability to observed scale. Heritability estimates ranged from 0.16 to 0.22 on the liability scale and 0.07 to 0.09 on the observed scale. Threshold models showed superior predictive ability compared to linear models, with higher accuracies (0.531 to 0.698 vs. 0.451 to 0.532), lower bias (−0.0004 to 0.008 vs. 0.027 to 0.096) and dispersion values closer to the ideal (0.932 to 1.000 vs. 0.811 to 0.848). Among the definitions, STAY48 with at least two or three calvings demonstrated the most consistent performance, representing a promising criterion for genetic evaluation in Nellore cattle.