Which experiment group is an imputed SNP associated with?

Question

[nathansoakes]

When conducting a GWAS, how do you interpret if an imputed SNP with a significant p-value and r2 is associated with the experimental group or control group? The following is what my intuition is telling me but it doesn't feel very rigorous:

Say, for a single imputed SNP, collect the typed SNPs that are in LD with the imputed SNP. From those either just average the odds ratios and assign that to the imputed SNP, or, count the typed SNPs with odd ratio greater than 1 and count the SNPs with odds ratio less than 1 and use those counts as the imputed odds ratio, from which I infer association.

Any insights would be appreciated.

Answer 1

[karl.stamm]

In a case-control comparison, the final SNP is associated with the difference in status. It implies an equal and opposite effect in both groups.

If you did the imputation, you would have a most-likely genotype for everyone at that locus, and can use that in a contingency table.