Bayesian multivariate reanalysis of large genetic studies identifies many new associations

• DOI: 10.1371/journal.pgen.1008431
• Publication Date: 2019-10-09T17:34:09Z
• Authors: Michael C. Turchin; Matthew Stephens
• Keywords: Genome-wide Association Study; Univariate; Genetic Association; Replicate

Genome-wide association studies (GWAS) have now been conducted for hundreds of phenotypes relevance to human health. Many such GWAS involve multiple closely-related collected on the same samples. However, vast majority these analyzed using simple univariate analyses, which consider one phenotype at a time. This is despite fact that, least in simulation experiments, multivariate analyses shown be more powerful detecting associations. Here, we conduct 13 different publicly-available datasets that phenotypes. These data include large anthropometric traits (GIANT), plasma lipid (GlobalLipids), and red blood cell (HaemgenRBC). Our identify many new associations (433 total across studies), replicate when follow-up samples are available. Overall, our results demonstrate can help make effective use from both existing future GWAS.