PLINK gxe not reading covariate file

Question

[jennavive00]

I am trying to conduct a GxE in PLINK using a covariate file but PLINK doesn't seem to be recognizing the file. Can someone tell me what I'm doing wrong?

plink --bfile omni --gxe --covar omni.txt

The covariate file omni.txt has the same # of observations as the .fam file and contains FID, IID, Covar1 coded like so...

FID  IID  Covar1
##   ##    1, 2, or -9

The log says that it reads the .bim and .fam files but doesn't even mention the covariate file.

[chrchang523]

Hi,

Can you post or send me the .log file?

[jennavive00]

PLINK v1.90b3j 64-bit (2 Apr 2015)
6 arguments: --bfile OMNI --covar omni.txt --gxe
Hostname: BR1
Working directory: W:\AIDGWAS
Start time: Thu Jun 18 11:18:48 2015

Random number seed: 1434651528
12245 MB RAM detected; reserving 6122 MB for main workspace.
608811 variants loaded from .bim file.
5582 people (2927 males, 2655 females) loaded from .fam.
5582 phenotype values loaded from .fam.

End time: Thu Jun 18 11:18:49 2015

[chrchang523]

This looks like a bug in the command-line parser/dispatcher; I will try to post a fix later today.

[jennavive00]

Thank you! :)

[chrchang523]

Okay, I've investigated this; the bug is a missing error message. PLINK should have printed Error: --gxe requires a scalar phenotype. before exiting. (Today's development build will do that.)

This brings us back to your dataset: --gxe does not handle case/control phenotypes, so you will need to use another analysis method.

[jennavive00]

I'm a little confused. If it doesn't do case/control studies then what is it used for? It says here https://www.cog-genomics.org/plink2/order that --gxe is used for quantitative phenotype + case/control covariate association test

[chrchang523]

The covariate is case/control, but the main phenotype has to be continuous.