Weird Q-Q plot in regenie while obtaining binary trait GWAS

[swathihg]

Hi everyone,

I am facing issue with obtaining the p-values close to expected values. I tried doing it in plink2 and regenie. WIth plink2, I got good results which align well with expected p-value whereas with regenie I am unable to get same results.

But when I use --ignore-pred flag in regenie I did get fairly good Q-Q plot. Please find attached log files for plink and regenie step1, step2. Any suggestions on this would be helpful. Thank you.

Attached are log files and plots:

PLINK 2
sum_stats_dep_without_hide_covar.log

REGENIE
fit_bin_out_dep.log
step2_chr1.log

PLINK 2

REGENIE

REGENIE using --ignore-pred flag

[Ojami]

What happens if you set --pThresh 0.999999 and remove --approx flag?

[swathihg]

@Ojami Thanks for your comment. The results are same as using --pThresh 0.01 with --approx flag

[joellembatchou]

Hi,

I see for PLINK you are testing on the BED file but with REGENIE the imputed BGEN data is used. Can you send sumstats for the variants with the smallest p-value in REGENIE? And the corresponding sumstats for this same set of variants using (a) REGENIE with --ignore-pred; (b) PLINK?

Thanks