Why my GWAS p-value QQ-plot is far above diagonal

Question

[Celia L.]

Hi. I'm trying to run GWAS pipeline using plink, but the results I got look really off. The QQ-plot of the p-values is far above the diagonal.

The phenotype I used is the height. I'm pretty sure I followed the correct QC steps. The inclusion of covariates is probably also not the one causing this issue, since with and without covariates I got very similar QQ-plots which look wrong (attached below).

Does anyone have experience with this and know what's going wrong? Thanks!

Answer 1

[LChart]

Height is the quintessential quantitative (in the sense of "hyper-polygenic") trait, with vast numbers of associated variants.

If you make a QQ plot of the recent summary from the GIANT consortium it looks like this:

and zooming in on just the first 10 orders of magnitude:

so do not expect height to look "well-calibrated"; in effect no variant in the genome is null.

[i.sudbery]

Its a great example of the omnigenic model. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5536862/