Genomic Prediction in a Wild Mammal

Genomic prediction is a molecular quantitative genetics method that has revolutionised animal and plant breeding and has several applications in human personalised medicine. This method uses genetic markers (SNPs) across the whole genome to predict breeding values of individuals. One of the objective of genomic prediction is to predict the phenotypes of an individual, such that genetically ‘best’ individuals can be used in breeding programmes.   Aim of this study was to investigate the genomic prediction accuracy in the population of Soay sheep (Ovis aries) which lives in NW Scotland. Genotype information at 37237 informative autosomal SNP markers on the OvineSNP50 BeadChip was available for a total of 5805 sheep. Traits used in this study were foreleg, hindleg, weight, metacarpal and hornlength. The Bayesian mixture model (called bayesR) used for genomic prediction. Cross validation was used to estimate the accuracies. Each data set of the traits were randomly split into a training sample containing 90%, 50% and 5% of individuals and a validation sample containing the remaining ones. Prediction accuracies were calculated as correlation between observed and predicted phenotype in the validation sample. Heritabilities were 0.44 (foreleg), 0.44 (hindleg), 0.12 (weight), 0.67 (metacarpal) and 0.25 (hornlength). The accuracies were reported as 0.42 (foreleg), 0.39 (hindleg), 0.17 (weight), 0.60 (metacarpal) and 0.34 (hornlength) when 95% individuals were in training set. Prediction accuracies increases as size of training population increased. Our results reveal that various traits measured in Soay sheep can be predicted with sufficient accuracy to make it feasible to implement genomic selection.