We tested the open question of whether mitochondrial dysfunction is a primary cause of type 2 diabetes (T2D) as opposed to a secondary cause or an outcome of the disease. Using a genetic approach, we comprehensively analyzed common variant associations at the level of genes and gene sets, in search for multiple modest genetic effects on T2D pathogenesis in a set of nuclear regulators of mitochondrial activity, the oxidative phosphorylation (OXPHOS) genes, or the full known set of ∼1,000 nuclear-encoded mitochondrial genes (an estimated 85% of all mitochondrial genes). For this analysis, we developed a modified GSEA approach applied to genetic association data (p-values or z-scores), which we named MAGENTA. MAGENTA was especially designed to exploit the increased power of meta-analyses of multiple GWA studies. In the process we identified and adjusted for confounders on gene scores and gene set enrichment scores in the absence of genotype information. This method was rigorously tested and evaluated using real and simulated GWA data, and we demonstrate realistic scenarios in which this approach could identify significant set-wide association signal that is likely to be overlooked in individual SNP analysis.
