Estimation & Prediction

Alternative Genomic Relationship Matrices and ssGBLUP Models for Improving Multi-Breed Genomic Prediction Accuracy in South African Dairy Cattle

---

Abstract

Limitations on the number of genotyped animals in reference populations make it challenging to implement genomic selection, especially in developing countries. Therefore, multi-breed genomic predictions were seen as an option to increase the reference datasets by pooling genotypes from different breeds. However, this strategy results in low accuracy except for when breeds share close genetic relationships. The objective of this study was to evaluate the accuracy of alternative genomic relationship matrices (GRMs) and alternative single-step genomic best linear unbiased prediction (ssGBLUP) models as strategies to enhance multi-breed genomic prediction in South African dairy cattle. Phenotypic and pedigree data were obtained from the South African Integrated Registration and Genetic Information System, and the edited data contained 865,073 pedigree and 1,289,836 lactation records (305-day) for milk, protein, and fat yields. The genotypes included were 2,574 animals (1,428 Holstein and 1,146 Jersey) after quality control and were generated using the Illumina 50K SNP chip. Three-trait multi-breed ssGBLUP evaluations were performed using three GRMs: (1) the standard VanRaden (2008) matrix, (2) the Wientjes et al. (2017) matrix accounting for between-breed covariance, and (3) a zero-relatedness matrix (ssGBLUP_Zero) assuming no genetic connection between breeds. The alternative ssGBLUP model incorporated inbreeding and unknown parent groups (UPG), with modified blending (G0.90, A0.10), tuning (TG1), and scaling (τ = 0.90; ω = 0.60). The results show that treating breed as a fixed effect improved prediction accuracy across all traits. The standard ssGBLUP achieved accuracies of 0.20-0.27 for Holstein cattle, while ssGBLUP_Zero improved combined-breed accuracies to 0.26-0.27. The alternative ssGBLUP model yielded the highest accuracies of 0.33 for milk, 0.38 for protein, and 0.39 for fat yield when validated with Holstein, which is a 10-16 % gain over the standard ssGBLUP. In conclusion, this study demonstrates that alternative GRMs and ssGBLUP models can substantially improve the accuracy of multi-breed genomic predictions in small reference populations. This study provides a practical strategy for achieving more reliable genomic evaluations in developing countries with limited reference populations.

Keywords: 2026

How to Cite:

Mafolo, K., Neser, F., Cason, E., MacNeil, M. & Makgahlela, M., (2026) “Alternative Genomic Relationship Matrices and ssGBLUP Models for Improving Multi-Breed Genomic Prediction Accuracy in South African Dairy Cattle”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2285907. doi: https://doi.org/10.31274/wcgalp.23820

Rights: 1