Gene-based testing was conducted using the primary discovery data in Europeans. Analysis was performed using the SNV-set Kernel Association (SKAT) test within the seqMeta R Package. SKAT tests were performed according to two different MAF filters of 1% and of 5%, and three different levels of variant filtering, based on annotations within the CHARGE Exome SNP Info annotation file: (i) all variants, (ii) variants deemed predicted to be damaging (24) and (iii) variants that were non-synonymous or leading to abnormal splicing. For gene-based tests we adjusted for multiple testing using the Bonferroni correction, according to the number of genes tested. The gene-wide significance level was calculated as 1.98 × 10−6 for 25 241 tests (i.e. the number of genes on the Exome Chip). For any genes attaining significance, the gene-based tests were repeated with exclusion of the most significantly associated lead variant, in order to confirm that the association was due to multiple rare variants.
