P.I. G5 Génétique Statistique « InBio » (C3BI)
Genome-wide association studies (GWAS) have proven successful in identifying thousands of significant genetic associations for multiple traits and diseases. This success is largely thanks to the dramatic increase in sample size achieved by GWAS meta-analysis consortia. Conversely, GWAS meta-analyses across different diseases and traits have received limited attention, even though multivariate analysis enables the detection of pleiotropic genetic variants. One important reason is that existing approaches require merging individual level data, a practically daunting and risky task in consortia including dozen or even hundreds of studies. To address this problem we propose JASS (Joint Analysis of Summary Statistics), a computationally efficient framework for the multivariate analysis of multiple GWAS summary statistics. Our framework solves all practical and methodological issues related to the analysis of aggregated statistics, while maintaining the gain in power expected for individual-level data. Applying JASS for the joint analysis of publicly available GWAS of multiple traits and diseases, we identified large number of genome-wide significant variants that were missed by univariate phenotype screening.