I am working with gwas summary statistics in order to prioritize novel candidate variants and genes.
For one locus, I found a strong association below the genome-wide significance threshold at 1x10-7. Next, i wanted to further prioritize the variant by eqtl/sqtl analysis with GTEx. However neither the lead SNP, nor the SNPs in LD > 0.7. Showed any functional annotation. After that I took a deeper look and found at the same locus variants (LD ~ 0.5, p-value 5x10-5), which have strong splice-qtl with a gene that interests me very much. By classical "gwas theory" I think I have a problem here, as they are in weak LD with the lead SNP. However, I figured out, that the 5x10-5 variants (AF = 0.2) have a D Prime of 1 with the 1x10-7 variant (AF = 0.4).
Could the following scenario take place here: Always when the rarer (causal) variants at my locus are present, the more common variants appear too. However, the more common ones appear in some more cases (just because their higher allele frequency) and thus get a higher p-value?