I have a question about imputation in GWAS. Since we often genotype only a subset of variants, for example around 500,000 SNPs, and then impute millions of additional variants using a reference panel, many association signals may come from imputed rather than directly genotyped variants. In general, do the majority of GWAS hits coming from the imputed variants??? How should we think about this bias based on imputed panel (with a particular population ex: HRC for European) when interpreting GWAS hits? Would it be fair to say that GWAS findings are partly conditional on the imputation reference panel and pipeline used? I don’t mean that imputed findings are unreliable, but rather that the confidence in an imputed GWAS hit depends on factors such as INFO/R² score, reference-panel ancestry match, allele frequency, and replication in independent samples. Is that the correct way to interpret it? Finally, Could the sample size of the imputed panel also plays a role including the ethnicities as well??