Estimation & Prediction

Multi-breed genomic prediction for feed efficiency using Nellore functional variants in Guzerat beef cattle

---

Abstract

In small tropical cattle populations, such as the Guzerat breed in Brazil, limited data hinder genomic evaluations for feed efficiency (FE). Multi-breed (MB) models leveraging larger reference populations, like Nellore, can enhance prediction accuracy, particularly when using the metafounder (MF) approach to address base-population mismatches. Additionally, incorporating functional information enhances the biological relevance of marker effects and may further improve genomic prediction (GP). This study aimed to evaluate the GP of MB model and the impact of integrating functional information for dry matter intake (DMI), residual feed intake (RFI), and residual gain (RG) in Guzerat cattle using Nellore as the reference population. Data were provided by the National Association of Breeders and Researchers (ANCP) in Brazil. The Nellore breed dataset included 18,567 animals genotyped with medium-panel and imputed whole-genome sequence data (1,547,715 SNPs after quality control) and 53,967 animals with pedigree information, of which 30,810 animals were phenotyped for DMI, RFI, and RG between 2010 and 2023. The Guzerat beef breed dataset comprised 1,501 genotyped animals (GGP Bovine 50K, Neogen) and 6,924 animals with pedigree information, including 1,322 animals with FE-related phenotypes measured, including DMI, RFI, and RG over the same period. Weighted single-step GWAS analyses were performed in the Nellore population using BLUPF90 programs to identify genomic regions associated with FE traits. A total of 21 genomic windows, each explaining more than 0.5% of the genetic variance, were annotated. Additionally, candidate functional variants (FV), including SNPs and INDELs, were identified via RNA-Sequencing of high- and low-RFI Nellore animals. The GP were evaluated in single-breed (SB) and MB scenarios, the latter using a MF approach to model breed compatibility. Three methods were applied in each scenario: standard single-step GBLUP, Weighted single-step GBLUP (WssGBLUP), and WssGBLUP using FV. These models were validated using the linear regression method on 119 young Guzerat animals born between 2022 and 2023, comparing whole and partial datasets to assess prediction accuracy, bias, and dispersion. Accuracy substantially improved in MB compared with SB models. Despite trait-specific differences, an overall negative bias was observed, indicating a mild underestimation of GEBV. Dispersion estimates improved in MB scenarios, indicating that MB approaches better manage prediction variance. Incorporating FV generally did not alter predictions. However, including FV for MB model increased prediction accuracy for RG (from 0.54 to 0.56) and reduced bias. This suggests that biologically informed variants may be particularly beneficial for improving genomic prediction of low-heritability traits like RG (0.14). Overall, MB models consistently improved GP accuracy and stability compared to SB approaches. While FV had a limited general impact, the specific improvement observed for RG suggests that biologically informed variants are most beneficial for low-heritability traits, reinforcing the effectiveness of MB strategies for related beef populations.

Keywords: 2026

How to Cite:

Teodoro, M., Gubiani, G., Londoño-Gil, M., Maiorano, A., Mota, L., Cánovas, Á. & Baldi, F., (2026) “Multi-breed genomic prediction for feed efficiency using Nellore functional variants in Guzerat beef cattle”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286857. doi: https://doi.org/10.31274/wcgalp.24147

Rights: 1