Improving genomic prediction in pigs by integrating multi-population data and prior knowledge

Genomic selection (GS) has become an essential tool for improving economically important traits in pigs. However, its accuracy depends heavily on the size and composition of the reference population. This study explores strategies for optimizing multi-population genomic evaluations by integrating prior biological knowledge and leveraging advanced genomic models. We assessed population similarities based on phenotypic distribution, linkage disequilibrium (LD) consistency, heritability, and genetic variance. Three genomic prediction models-GBLUP, bivariate GBLUP, and GFBLUP-were applied to evaluate the joint reference populations. The results indicated that differences in phenotypic means and genetic variance between populations significantly affected the prediction accuracy of joint evaluations, particularly for fat thickness traits. The GFBLUP model, integrating meta-GWAS priors, improved prediction accuracy when the genetic contributions were similar between target and reference populations. These findings highlight the importance of carefully selecting reference populations and integrating biological priors into genomic evaluations. The study offers valuable insights for optimizing genomic selection strategies in pig breeding programs.