We applied two approaches to gene- and set-based analyses, FUMA (Watanabe et al., 2017) and JEPEGMIX (Lee et al., 2016). The former conducts gene-based tests across all markers followed by gene-set analysis. We submitted meta-analysis summary statistics to the FUMA pipeline, which requires selection of a reference panel in order to account for linkage disequilibrium (LD) among markers. We selected the EUR subsample of the 1000 Genomes reference panel as this group constituted >82% of the sample, and correcting for EUR LD is a more conservative approach than correcting for AFR LD (AFR being the next-largest component of the full sample). JEPEGMIX differs from FUMA in that its gene- and set-based analyses are tissue-specific: using GWAS summary statistics, it tests the joint effect of functional SNPs known to affect the expression of a gene, effectively predicting whether tissue-specific gene expression is associated with an outcome of interest (here, SRE score). The method can be extended to estimate the joint effects across gene sets.
