Here we introduce a GSEA approach applied to genome-wide variant association data, which we named “Meta-Analysis Gene-set Enrichment of variaNT Associations” (MAGENTA). MAGENTA does not require genotype data, making it especially relevant to GWA study meta-analyses. We tested and validated MAGENTA using the Diabetes Genetics Initiative (DGI) GWA study [17], and three GWA meta-analyses of cholesterol and lipid blood levels [34]. Using simulations, we identified the conditions under which our method has increased power to detect associations for which there is low detection power with single SNP analysis. Finally, to test whether mitochondrial dysfunction may be causal to T2D, we applied MAGENTA to a set of known nuclear regulators of mitochondrial genes [35], the OXPHOS genes [9], and all known (∼1,000) autosomal human mitochondrial genes [22], using the latest T2D meta-analysis of a total of 47,117 individuals (DIAGRAM+) [36], as well as meta-analyses (up to 46,186 individuals) of seven glucose and insulin-related traits relevant to T2D pathogenesis (MAGIC; [37], [38], Soranzo N. et al., unpublished data).
