Computational protocol: Multivariate simulation framework reveals performance of multi-trait GWAS methods

Similar protocols

Protocol publication

[…] In this comparison study we compare single-SNP multi-trait methods, both those that use individual-level genotype-phenotype data: MANOVA, CCA (mv-PLINK), Combined-PC, MultiPhen, mv-BIMBAM and mv-SNPTEST, and those that exploit GWAS summary statistics: min-P9, TATES, SHom, and SHet. These methods cover several approaches to testing the association of genetic variants with multiple phenotypes, including multiple linear regression techniques, a reversed (ordinal) regression with SNP as outcome, simple and complex adjustments of summary statistic results, across trait meta-analysis techniques and Bayesian methods (see Methods). provides a summary of the methods compared in this study. The simulation framework provides a thorough and consistent platform on which to compare these different methods. We illustrate power across the full range of phenotypic correlations (e.g. ), for up to 48 phenotypes, which we believe represents the clearest way to expose differences in method performance. […]

Pipeline specifications

Software tools PLINK, MultiPhen, BIMBAM, SNPTEST, TATES
Application GWAS