Meta-Analyses Approaches to Prioritise Pleiotropic Sequence Variants from Global Sequence Consortia

Run6 of the 1000 Bull Genomes Consortium data contains approximately 2200 Bos Taurus animals and 400 Bos indicus and admixtures, ancient cattle including an auroch, and other closely related Bos species and will result in >67 million variants. Similarly, Run2 of SheepGenomesDB contains 935 diverse sheep yielding >50 million variants. Genomic prediction methods struggle with millions of sequence variants. Therefore we need to prioritise sequence variants using statistical and functional methods to harness their benefit on genomic prediction accuracy. A prioritised variant set of say 100,000 variants would have to be unified across several important traits for routine genetic evaluation, which makes pleiotropic variants an important early focus. Approximate multi-trait methods that perform meta-analyses of single-trait results are more tractable than full multi-trait genomic prediction. We developed three such methods to map pleiotropic quantitative trait loci (QTL) and tested them in large sheep meat and wool data sets of up to 10,613 animals imputed for 510,174 SNP. One method was a multi-trait genome-wide association study (multi-GWAS), while the other two methods utilised BayesR genomic prediction outputs, such as local GEBVs and posterior probabilities of having an effect. In the 56 carcass composition traits, multi-GWAS identified putative pleiotropic genes in several functional groups. 206 pleiotropic QTL were discovered in the 44 wool traits. While the two meta-approaches using Bayesian statistics provided a more precise mapping of QTL than multi-GWAS. The next step is to apply these methods to sequence variants and evaluate their impact on the accuracy of genomic prediction.