Genomic estimation of dominance variance and inbreeding depression in a local sheep breed

Dominance variance can account for a substantial proportion of total genetic variance depending on the species and trait. Its assessment was historically cumbersome primarily due to the size and structure of available pedigree data and corresponding computational complexities. Estimations were simplified with the availability of genome-wide SNP markers, opening the potential to study non-additive genetic variation even in small populations such as local sheep breeds. Further, it has been argued that when estimating dominance, it is essential to account for genomic inbreeding because a decrease in heterozygosity can lead to an overall decrease in the mean of a trait of interest due to directional dominance. Local livestock breeds are susceptible to inbreeding and inbreeding depression due to small population sizes and the resulting increased relatedness between mates. Local breed conservation should be prioritised because they are reservoirs of genetic diversity and have characteristics that could be essential resources for adapting to future challenges. Therefore, when designing selection programmes for small livestock populations, the effects of dominance and inbreeding need to be considered. We aimed to estimate additive and dominance genetic variances and genomic inbreeding for milk traits in Pag sheep. This local Croatian breed is adapted to a Mediterranean island with a marginal grazing system and is raised for milk (mainly used for cheese) and lamb production. We had 50K SNP array genotype data for 2134 animals, of which 1744 were ewes with milk records. After quality control of genomic data, we imputed sporadic missingness and corrected genotype errors using AlphaPeel. We detected runs of homozygosity (ROH) with Plink 1.9 to estimate genomic inbreeding (FROH). For each recorded milk trait (milk, fat and protein yields in kg, and somatic cell score), we compared four single-trait models with additive and dominance effects, and FROH as a covariate to account for directional dominance, fitted with both Bayesian and REML methods as implemented in BLUPF90. We found additive variance was stable across the different models for each trait, while dominance variance varied and was impacted by the inclusion of FROH covariate in the model. Dominance variance accounted for 10-30% of genetic variance across models and traits. Using a genome-wide association analysis approach, we will also detect regions of the genome with additive genetic, dominance genetic or ROH effects on milk traits. Finding the best approach to utilise available non-additive genetic variation in the local (small) livestock breeding programmes via either optimal contribution selection and mate allocation schemes or leveraging SNP effects associated with the non-additive genetic effects is an important open question. Genomic data has allowed us to estimate additive and dominance variance and inbreeding in Pag sheep, which will contribute to developing sustainable genomic selection programmes for this and other small livestock populations.