GAWMerge has some limitations. First, careful consideration of not only ancestry, sex, and age distributions, but other systematic differences between a given case-only cohort and public controls, like smoking status, is essential to unbiased use of public controls and application of GAWMerge. All association analysis conducted included ten principal components as covariates to account for population substructure, although applying GWAS in as homogenous population as possible is desirable. This requirement places some limits on the public controls that can be used for any given case-only cohort. Second, the additional QC steps might mask some real trait-associated variants. In the attempt to recover the known genetic variants associated with COPD, there were two loci (RIN3 and MMP3/12) not reaching the genome-wide significance in the meta-analysis (Table 2). The three SNPs were filtered out in the first GWAS, comparing COPD cases in COPDGene EA with WGS data and smoking controls in COGEND EA with array data, due to high R2 difference between the WGS and array data. Thus, GAWMerge may lose some sensitivity while controlling type-I errors. There is also the potential
