In order to determine whether the validated secondary signals at previously reported loci were independent of the published SNV, conditional analysis was performed within GCTA software (34) applying the –cojo method (consisting of conditional and joint analysis with stepwise model selection). The input data were the exome-wide summary statistics from the full discovery meta-analysis of RR-interval in Europeans. The 1958 Birth Cohort Study (1958BC; N = 5815) dataset was used as the reference for genotype data, because it represents one of the largest discovery studies (See Supplementary Material, Table S13). LD was calculated between pairwise SNVs, but any SNVs further than 10 Mb apart were assumed to not be in LD. All autosomal chromosomes were analysed, with MAF restricted to ≥ 0.01%, to allow for low frequency secondary SNVs, whilst taking into account the statistical power achievable. To allow for secondary associations a P-value cut-off of 1 × 10−4 was used as the modelling selection threshold within the GCTA analysis. Results were then extracted for the nine previously reported regions, within which potential secondary signals had been validated from the
