Closed ameynert closed 3 years ago
This appears to be the same issue #3328. I've swapped in the 1.2.0 vt and vcfstream executables which I have on another system entirely, and they work fine.
Thank you so much @ameynert, sorry for being slow in responding. Let me see if I can reproduce this-- we can pin to 1.2.0 but I'd rather either get 1.2.3 fixed or otherwise handle it.
Thanks @roryk. I've managed to run the ggd recipe standalone with the swapped in version of vt, and copied the output into hg38/variation. Is there anything else that's required for the gnomad datatarget or can I resume my bcbio installation excluding '--datatarget gnomad'?
Thanks, that should work, yes. Thanks so much for figuring out what's going on. I'm pretty sure this problem is related to some htslib 1.9 and htslib 1.10 incompatibility issues, bioconda is kind of in an in-between state right now and some tools are broken on specific versions, working on it now.
Thanks, I can't seem to reproduce this-- could you let me know what verision of samtools
is installed?
Version: 1.10 (using htslib 1.10.2)
It's a clean system with only the bcbio install, no other samtools/htslib installed.
Thanks, what about bcbio_conda list vt
what does that show?
Ok, I pinned our bcbio install to samtools 1.9, most of bioconda is built against htslib 1.9 right now-- we have a separate htslib1.10 environment for tools/updates that depend on it so that should fix problems with htslib, but I couldn't make this bug happen with 1.9 or 1.10.
Ok, it was a htslib1.10 issue. I ended up being able to reproduce it. If you do a
bcbio_nextgen.py upgrade -u development --tools
it will pull in samtools 1.9, and this issue should be resolved. Thank you!
I've traced the error to vt normalize in this line of ggd-run.sh:
I pulled out the first 10k lines of the downloaded gnomAD genomes VCF for testing:
Full bcbio install log from relevant point: