I have just realized that a GWAS that I ran sadly included related individuals (the flagged samples file that came with the dataset did not include relatedness). As relatedness is expected to increase false positives, is it reasonable to solely rerun the GWAS on the genome-significant hits, or do I need to rerun the entire analysis? We are mostly concerned about false positives with the results.
I filtered out samples that had extreme principal components values, which should have helped capture some of this. Only around 255 samples were removed out of over 10k)
Regenie is supposed to help control for relatedness, so we'll see how removing related individuals could affect things.
You can use a method which explicitly models relatedness in the samples, so you can don't need to exclude them. We have found that BOLT-LMM works well for this if you are analysing a quantitative (continuos) trait.
Id just re-run the entire analysis. its not going to take that long in the scheme of things. just keep writing the discussion or what have you. Your results are unlikely to change significantly, so just keep plugging away, but do re-run it from the top.
While you are at it, make sure your other SNP level and Sample level QC metrics are in order, too!
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Does this answer your top level question or add additional information/context to it?
It should help control some bias and perhaps touch on related individuals, but I am seeking guidance/information from others.