Combining association signals for multiple loci across genes and gene-sets has the potential to capture a greater proportion of variance leading to an increase in power [35]. We performed gene-based association analyses, using the VEGAS2 method (performed at: https://vegas2.qimrberghofer.edu.au/) [36]. This method calculates a test statistic from the sum of test statistics within a gene region. The LD between markers within a gene region is calculated and adjusted for within the software using the 1000 genomes reference genotypes. Analyses were limited to the top 10%, by P value, of SNPs per gene, an approach which has been shown to give rise to higher sensitivity and lower false positive rates compared to other gene-set methods [37].
