Conservation & local programs

Using a large commercial reference population to predict performance in a local pig breeding program

---

Abstract

Local breeding programs often struggle to achieve fast genetic progress due to their limited reference populations. In contrast, global commercial breeding programs have developed extensive reference populations, enabling faster genetic progress. Australian pig breeding programs provide an interesting case study, as the Australian market has been closed to external genetics (e.g., live animals, semen, embryos) for more than 20 years. Despite this prolonged isolation, Australian pig populations and other local breeds still share ancestral haplotypes with populations from global breeding programs, suggesting potential benefits from joint international genomic evaluations. In this study, we investigated genomic prediction accuracy for lifetime daily gain (LDG) and backfat (BF) in pigs from an Australian breeding program. We compared evaluations based solely on the Australian reference population, solely on an international commercial reference population, and a joint genomic evaluation combining both populations. The Australian dataset comprised approximately 36,000 synthetic-line animals, while the commercial dataset included approximately 240,000 purebred Duroc animals, all phenotyped for both traits. The genetic background of the synthetic line is not well documented, but it includes Duroc. In total, 1,670 Australian animals and 81,200 commercial animals were genotyped using the Illumina 50K SNP chip. Only SNPs segregating in both populations were analyzed. The youngest 835 genotyped animals from the Australian synthetic population were designated as the validation population. Lifetime daily gain was defined as average daily gain from birth to approximately 120 kg, and backfat was measured at approximately 120 kg. Genomic predictions were obtained using a single-step genomic evaluation implemented in MiXBLUP. For both traits, the statistical model included fixed effects of sex and herd-year-week, and random effects of additive genetic merit, common litter, pen, and residual error. For BF, body weight at test was included as a covariate. Prediction accuracy was calculated as the correlation between estimated breeding values and phenotypes of validation animals after correction for all non-genetic effects. Phenotypes were pre-corrected using the full Australian dataset. The highest prediction accuracies were obtained when the Australian synthetic reference population was used alone, with values of 0.22 for LDG and 0.26 for BF. Using only the commercial Duroc reference population resulted in lower accuracies, with both traits reaching 0.14. The joint genomic evaluation produced accuracies similar to those obtained using the Australian reference population alone, with values of 0.22 for LDG and 0.24 for BF. These results indicate that although the commercial population contains informative ancestral haplotypes, its predictive ability for the genetically isolated Australian population is limited compared with locally derived reference data. Nevertheless, the findings demonstrate that international commercial genomic resources can still provide value to smaller breeding programs with restricted data, highlighting the growing potential of genomic globalization to overcome geographical and historical barriers.

Keywords: 2026

How to Cite:

Lopes, M., van der Spek, D., Bunz, A., Vargovic, L., Grindflek, E. & Harper, J., (2026) “Using a large commercial reference population to predict performance in a local pig breeding program”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286468. doi: https://doi.org/10.31274/wcgalp.24019

Rights: 1