For GWAS summary statistics, can I calculate MAF using p valuee, OR and N?

Question

[YL]

Hi,

I am interested in using Coloc which requires MAF as input. However, my GWAS summary statistics only have p values, OR and N (including N of cases and controls). I wonder if it would be possible to calculate MAF, and how to do so?

Thank you very much!

Answer 1

[LChart]

You can't do this because of uncertainties between the standard error (which partially depends on unknown correlations between genotype and covariates) and frequency.

However you probably do have the variant ID (rsID or position) -- otherwise how could you hope to perform coloc -- so you can look up the variant frequency in dbGAP. If you need the MAF in the cohort, you can use disease prevalence, population MAF, and the odds ratio to compute the prevalence in cases and controls (https://heavytailed.wordpress.com/2013/06/).

Answer 2

[YL]

Thank you very much! I understand that I could use population MAF from a reference dataset as a proxy

For Coloc, they can accommodate other inputs including using both beta and variance of beta. I ended up calculating variance of beta by solving this formula pvalues=pnorm(-abs(beta/sqrt(variance of beta)))*2 (https://chr1swallace.github.io/coloc/articles/a02_data.html#what-if-i-don-t-have-beta-and-or-varbeta-)