Different genotyping chips from Illumina (Illumina Inc., San Diego, CA) were used (table 1). QC steps were previously described in detail for the NETT, Norwegian and ECLIPSE cohorts,23 and were applied to the COPDGene study as well. Briefly, QC steps consisted of filtering SNPs based on missing call rates (>5%) and Hardy-Weinberg Equilibrium deviation (p<10−8), and filtering subjects based on genotyping call rate (<95%), sex discrepancy, unexpected relatedness (PLINK pi-hat cutoff of 0.125), and ethnicity. Removal of cases that were outliers for genetic ancestry was performed based on principal components (PCs) analysis in cases only. In the primary analysis, untyped markers were imputed using 120 founder Caucasian (Centre d’Etude du Polymorphisme Humain (Utah residents with ancestry from northern and western Europe) (CEU)) haplotypes from HapMap reference panel (phase II) and, as a secondary analysis, using 1000 Genomes Project data.2425 We limited our analysis to SNPs genotyped/imputed in at least 2 cohorts, with imputation r2 coefficient ≥0.3 (for imputed SNPs only). Overall, approximately 2.5 and 6.3 million SNPs per phenotype were analyzed using the reference HapMap Project and 1000 Genomes Project panels, respectively.
