As UK biobank genetic data were released in two phases, we took the opportunity to replicate findings from the discovery stage in a further 275,596 individuals made available in the phase two release of UK Biobank genetic data. To avoid potential relatedness between discovery and replication samples, the replication samples were screened and individuals with relatedness closer than second degree with the discovery sample in the UK Biobank were removed [24]. Phenotypes were defined in the same way as the discovery samples (described above). Since the exome array and the UK Biobank Axiom arrays do not fully overlap, we used both genotyped exome variants (approx. 64,000) as well as the additional ~90,000 well-imputed exome array variants from UK Biobank (imputation quality score >0.3) for replication of single variant and gene-based tests. The rare ATF6 variant was absent from the UK Biobank array and is more prevalent in Africans (MAF = 0.01) than Europeans (MAF = 0.0007). Therefore, replication was sought in 1,437 individuals of African American-ancestry from the HRS and COGA studies. Analysis methods for replication cohorts were the same
