Estimation & Prediction

Enhancing Genomic Prediction Accuracy for Breeds with Numerically Small Reference Population Using Multi-Breed Evaluation Strategies

---

Abstract

Genomic selection has accelerated genetic improvement in cattle; One of the major factors affecting accuracy of genomic selection is the size of reference population. However, its implementation in developing countries remains constrained for breeds with numerically small reference population, which are often limited by the lesser number of animals with both genotype and phenotype available. The present study explored the effectiveness of multi-breed reference populations for improving genomic prediction accuracy in indigenous cattle breeds with limited reference population under resource-scarce conditions. The dataset included 305-day first lactation milk yield (305-DMY) records and SNP genotypes from 1,298 Gir, 1,291 Sahiwal, and 500 Kankrej animals. Principal component analysis combined with K-means clustering revealed partial genetic overlap between Gir and Kankrej, supported by similar linkage disequilibrium decay trends, indicating shared haplotype structure. Heritability estimates for 305-DMY were 0.30 ± 0.07 in Gir, 0.27 ± 0.07 in Sahiwal, and 0.17 ± 0.01 in Kankrej. Genomic estimated breeding values were obtained using GBLUP with four evaluation strategies: shared GRM, non-shared GRM, metafounder-corrected shared GRM, and multi-breed multi-trait GBLUP. Single-breed evaluations yielded prediction accuracies of 0.65 (Gir), 0.60 (Sahiwal), and 0.49 (Kankrej). Multi-breed genomic prediction improved accuracy for Kankrej substantially by 23.6%, 24.6% and16.9% under shared GRM, non-shared GRM, and metafounder-corrected shared GRM approaches, respectively with combined Gir-Kankrej reference population, while there was no change observed with muti-breed multi-trait model. In contrast, Gir showed decline (1 to10%) in accuracy under multi-breed genomic evaluation, while Sahiwal exhibited variable responses, including a modest increase when combined with Gir (1 to 8%) and a decrease when combined with Kankrej (2-10%).Validation using the linear regression (LR) method demonstrated that the benefit of employing a multi-breed reference population was most evident for Kankrej, the breed with numerically small reference population, which exhibited the greatest improvement in genomic prediction accuracy. The highest gains for Kankrej were obtained when it was evaluated using Gir-Kankrej combined reference populations, with the shared GRM approach producing the largest increase in accuracy. These results indicate that multi-breed genomic evaluations can be effective strategy to enhance genomic prediction accuracy for breeds with small reference populations when genetic similarity exists between contributing breeds; however, it depends on breed combinations and genomic relatedness.

Keywords: 2026

How to Cite:

ALEX, R., Dawood Khan, K., Gowane, G., Nayee, N., Paul, R., V.N., S., VOHRA, V. & Yadav, A., (2026) “Enhancing Genomic Prediction Accuracy for Breeds with Numerically Small Reference Population Using Multi-Breed Evaluation Strategies”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286370. doi: https://doi.org/10.31274/wcgalp.23964

Rights: 1