--pca ending in Segmentation Fault (core dump)

169 views
Skip to first unread message

Letitia Sng

unread,
Jan 7, 2021, 5:23:48 PM1/7/21
to plink2-users
Hi, I am trying to calculate PCs of the 1000 Genomes using the following command:

plink2 \
    --vcf chrALL.1000g.ukbb.recode.vcf.gz \
    --extract chrALL.1000g.ukbb.prune.in \
    --freq counts \
    --pca allele-wts vzs \
    --out chrALL.1000g.ukbb.pruned \
    --memory 128000 \
    --threads 15 \
    --debug

But once the GRM has been constructed, I get a segmentation fault with a core dump, ending my run prematurely. 

I have been running Plink version: PLINK v2.00a3 AVX2 (28 Oct 2020). I've also included the log file. Any help would be greatly appreciated! 
chrALL.1000g.ukbb_hypertension.pruned.log

Christopher Chang

unread,
Jan 7, 2021, 7:12:57 PM1/7/21
to plink2-users
This is definitely a bug, but I haven't been able to replicate it.  What operating system are you running on?  Is it possible for you to send me a set of input files that I can use to try to replicate the crash?

Letitia Sng

unread,
Jan 7, 2021, 8:12:24 PM1/7/21
to plink2-users
It's running on:

Distributor ID: SUSE
Description:    SUSE Linux Enterprise Server 12 SP4
Release:        12.4

The (.gz) input files are quite large to send, but I could upload it somewhere. It's the publicly available 1000 Genomes Phase3 Autosomal vcfs that have been filtered to common variant locations. 

Christopher Chang

unread,
Jan 7, 2021, 8:34:27 PM1/7/21
to plink2-users
Ok, it should be sufficient to post the output of --write-snplist on the original VCF, and the .prune.in file, then.

Do you know how this plink2 build was compiled?  Its version string does not correspond to the Linux precompiled binary (no "Intel"), so I should try to link to the same linear algebra library, etc.

Letitia Sng

unread,
Jan 7, 2021, 9:58:33 PM1/7/21
to plink2-users
A .zip of the output of --write-snplist and the prune.in file can be found here

This job is being run on my organisation's HPC, so I don't have the details on how the build was compile, but I can find that out. It might take some time though.

Christopher Chang

unread,
Jan 8, 2021, 2:41:13 AM1/8/21
to plink2-users
Ok, let me know when you have that information, I'm still unable to reproduce the segfault.

Letitia Sng

unread,
Jan 10, 2021, 5:08:37 PM1/10/21
to plink2-users
It looks like this issue was due to the compilation on the cluster. The commands now work with no segmentation fault. Thanks for help and quick responses!
Reply all
Reply to author
Forward
0 new messages