Using EPACTS to implement SKAT-O - question about ped files

Question

[LauferVA]

Hello,

I am using EPACTS to implement SKAT-O for gene based burden testing. To do this, I have a multisample VCF and I have created a grpfile and a pedfile, according to EPACTS documentation.

However, I set the phenotypes to -9 in the pedfile, but I do not believe they are being excluded... has anyone used EPACTS and SKAT together? If so, can you tell me if the program is running the same PLINK would, in your experience? Within EPACTS, the context is once again to run a gene based test using SKAT-O but without 60 of the samples in the multisample test.

If worst comes to worse, I could write a brand new VCF without the 60 samples in it, then create a new ped file without those individuals, but it would be a heck of a lot easier just to set them to missing and have EPACTS / SKAT ignore them.

To be clear, the reason why I do not think they are being excluded is that, in the output file (the .epacts file) it lists the N as 305 - it should be 243 with everything removed. In the .phe file, it lists the removed genomes as 0, but in the log files that my scripts generate, it also says:

Detected phenotypes with more than 3 unique values -- considering as quantitative phenotypes. Successfully written phenotypes and 9 covariates across 305 individuals

To me this indicates the data for these individuals might be used even though they should be excluded...

Any help would be much appreciated. Thanks.

Answer 1

[dheeraj.bobbili]

Hi,

If you want to exclude any individual from the analysis just remove them from the PED file because EPACTS only uses the overlapping ids from VCF and PED file.

Hope it helps.

Answer 2

[karl.stamm]

A lot of tools don't handle missing values well.

To make a 'cleaned up' PED, you can drop rows with the -9 with this simple grep command

cat IN.PED | grep -v "-9" > OUT.PED