'Meta-founders' to Model Base Populations in Genomic Evaluation for Multi-Breed Sheep Data

Abstract

Models for genetic evaluation of animals from different base populations need to account for systematic differences not explained by genetic relationships considered. These include differences between breeds, animals with unknown parentage born in different time periods or, for single-step evaluation, founders for animals with or without genotype information. A standard method to achieve this, is to define appropriate genetic groups and fit these as additional effects in the model of analysis. Recently, so-called meta-founders have been proposed as an alternative which accounts for ancestral inbreeding and relationships, etimated from genomic information. We examine estimates of ancestral relationships and their impact on predicted breeding values for a practical data set from a multi-breed sheep population. While estimates were afflicted by insufficient genomic information for some groups, results correctly identified some known breed of strain differences and patterns of introgression. Correlations between predicted breeding values from respective analyses fitting genetic groups and meta-founders were high, suggesting that there is scope for meta-founders to replace genetic groups. However, fitting meta-founders reduced variances of predicted breeding values. Gurther investigations when more genotype information becomes available are warranted.