I am interested in looking for genes that are disproportionately affected by mutations in cases compared to controls. Originally, the analysis that was done was hand-picking affected genes (based on SNPs that came out from a GWAS) based on biological "interestingness" and just doing a 2x2 contigency table (cases, controls, and number of SNPs affecting that gene in each) followed by a Fisher's exact test.

Besides not performing a multiple test correction, is there something inherently wrong with this approach? Looking for a possible alternative to addressing this question ("do we find gene X consistently more affected by SNPs in cases than in controls?").