Non parametric quantitative genome wide association tests

Question

[Argus]

I am running a Genome Wide Association test, using a quantitative phenotype. This phenotype can have quite a spread with some extreme outliers, which invalidates the normality assumption for linear regression. I have tried doing various transformation to the phenotype data (log, power, ...) but still end up with a few individuals heavily weighting the results. I currently have my data in PLINK format.

I was wondering if there was a way to run a non parametric test, such as the Mann-Whitney U test?

Answer 1

[chrchang523]

One possible workaround is quantile-normalizing your phenotype (forcing it to a normal distribution, keeping its rank-order). PLINK 2.0 has the --quantile-normalize flag for this, and this should be straightforward to do in R/Python/etc. as well.