Leveraging large scale beef cattle genomic data to identify the architecture of polygenic selection and local adaptation

Since the invention of the first array-based genotyping assay for cattle in 2008, millions of animals have been genotyped worldwide. Leveraging these genotypes offers exciting opportunities to explore both basic and applied research questions. Commercial genotyping assays are of adequate variant density to perform well in prediction contexts but are not sufficient for mapping studies. Using reference panels made up of individuals genotyped at higher densities, we can statistically infer the missing variation of low-density assays through the process of imputation. Here, we explore the best practices for performing routine imputation in large commercially generated genomic datasets of U.S. beef cattle. We find that using a large multi-breed imputation reference maximizes accuracy, particularly for rare variants. Using three of these large, imputed datasets, we explore two major population genetics questions. First, we map polygenic selection in the bovine genome, using Generation Proxy Selection Mapping (GPSM). This identifies hundreds of regions of the genome actively under selection in cattle populations. Using a similar approach, we identify dozens of genomic variants associated with environments across the U.S., likely involved local adaptation. Understanding the genomic basis of local adaptation in cattle will enable select and breed cattle better suited to a changing climate.