We suspected that the false positives we observed derived from the phasing step since phasing of the array and WGS genotypes was based on different sets of variants. In addition, the TOPMed phased WGS data were derived from the samples of all studies26, which is different from the sample set we used, the COPDGene cohort, for phasing the array data. We repeated the technical comparison, using the same set of QC-validated variants and samples (Fig. 2a) as the basis for separate phasing of the array and WGS data, followed by the subsequent steps in GAWMerge (Fig. 1). The array data were specified as the case group for association testing, and the WGS data were specified as the control group, for European ancestry (EA) and African-ancestry (AA) separately. The results (Supplementary Fig. 4) confirmed that phasing based on a common set of variants and samples followed by the additional steps of GAWMerge eliminated false positives and made array and WGS data comparable for conducting GWAS.
