Run times for PCGR v2.1.2

[ridwan-shaikh]

Hi Sigve and Peter,

I was wondering if you have any bench marks for PCGR v2.1.2? We are currently updating from v1.4.1 but I've noticed the run times on the latest version is much longer. I've tried across two servers from 2 CPU cores to 8 CPU cores and with adjustments to the VEP buffer size and VEP forks but on average I am finding the run times to be approx. 24-28mins for a sample with 863 variants. Compare that to v1.4.1 which would take approx 3-4mins per sample.

Where I am finding the process to take the longest is VEP NearestExonJB.

2025-01-17 17:56:07 - pcgr-vep - INFO - VEP configuration - buffer size/number of forks: 500/4
2025-01-17 17:56:07 - pcgr-vep - INFO - VEP - plugins in use: NearestExonJB
2025-01-17 18:23:09 - pcgr-vep - INFO - Finished pcgr-vep

with debug mode on:

2025-01-17 18:05:22 - pcgr-vep - INFO - VEP configuration - buffer size/number of forks: 1000/8
2025-01-17 18:05:22 - pcgr-vep - INFO - VEP - plugins in use: NearestExonJB
2025-01-17 18:05:22 - pcgr-vep - INFO - unset PERL5LIB && export PATH=/opt/mambaforge/envs/pcgr/bin:"$PATH" && vep --input_file /mnt/data/4.testing/PCGR_update/SNVs/8_cores/RMH200SolidPromega_Pos-T.pcgr.grch37.fo1d6ffp69mymjn.ready.vcf.gz --output_file /mnt/data/4.testing/PCGR_update/SNVs/8_cores/RMH200SolidPromega_Pos-T.pcgr.grch37.fo1d6ffp69mymjn.vep.vcf --dir /mnt/software/7.resources/hg19/PCGR --assembly GRCh37 --cache_version 112 --fasta /mnt/software/7.resources/hg19/PCGR/20240927/data/grch37/misc/fasta/assembly/Homo_sapiens.GRCh37.dna.primary_assembly.fa.gz --pick_order mane_select,mane_plus_clinical,canonical,biotype,ccds,rank,tsl,appris,length --buffer_size 1000 --fork 8 --hgvs --af_gnomad --variant_class --domains --symbol --protein --ccds --mane --uniprot --appris --biotype --tsl --canonical --format vcf --cache --numbers --total_length --allele_number --failed 1 --no_stats --no_escape --xref_refseq --vcf --check_ref --dont_skip --flag_pick_allele_gene --plugin NearestExonJB,max_range=50000 --force_overwrite --species homo_sapiens --offline --quiet --regulatory
2025-01-17 18:29:00 - pcgr-vep - INFO - bgzip -f -c /mnt/data/4.testing/PCGR_update/SNVs/8_cores/RMH200SolidPromega_Pos-T.pcgr.grch37.fo1d6ffp69mymjn.vep.vcf > /mnt/data/4.testing/PCGR_update/SNVs/8_cores/RMH200SolidPromega_Pos-T.pcgr.grch37.fo1d6ffp69mymjn.vep.vcf.gz
2025-01-17 18:29:00 - pcgr-vep - INFO - tabix -f -p vcf /mnt/data/4.testing/PCGR_update/SNVs/8_cores/RMH200SolidPromega_Pos-T.pcgr.grch37.fo1d6ffp69mymjn.vep.vcf.gz
2025-01-17 18:29:00 - pcgr-vep - INFO - Finished pcgr-vep

I've had a look at the logs for our analysis on v1.4.1 and noticed that there was a precedence issue with a VEP perl module:

2025-01-16 23:52:14 - pcgr-vep - INFO - VEP configuration - buffer_size/number of forks: 100/4
perl: warning: Setting locale failed.
perl: warning: Please check that your locale settings:
        LANGUAGE = (unset),
        LC_ALL = (unset),
        LC_CTYPE = "C.UTF-8",
        LANG = "en_US.UTF-8"
    are supported and installed on your system.
perl: warning: Falling back to the standard locale ("C").
Possible precedence issue with control flow operator at /opt/mambaforge/envs/pcgr/lib/site_perl/5.26.2/Bio/DB/IndexedBase.pm line 805.
2025-01-16 23:52:23 - pcgr-vep - INFO - Finished pcgr-vep

Is this how long it takes for your samples?

Happy to share data to help.

Regards,
Ridwan

[sigven]

Hi Ridwan,

Thanks for reaching out! I'll have to look more carefully into that issue, I must admit I haven't experienced any similar (at least noteworthy) decrease in run time, so i'd have to do a bit of testing. Which setup are you using, Conda/Docker/Singularity? And are you able to share your test VCF with us?

kind regards,
Sigve

[ridwan-shaikh]

Hi Sigve,

I am using the singularity container butttt its actually the docker container converted to the singularity version, mainly because the oras link wouldn't work:

[ridwanshaikh@node01(alma) PCGR]$ apptainer pull oras://ghcr.io/sigven/pcgr:2.1.2.singularity
FATAL:   While pulling image from oci registry: error fetching image to cache: failed to get checksum for oras://ghcr.io/sigven/pcgr:2.1.2.singularity: unexpected docker media type received; try changing the protocol to docker://

so switching it to:

[ridwanshaikh@node01(alma) PCGR]$ apptainer pull docker://sigven/pcgr:2.1.2

works instead. It's the way I tend to create apptainer/singularity containers usually anyway. Docker -> DockerHub -> pull using apptainer. So I doubt that fact would make a difference, but happy to test the docker method only if that helps?

Yes, I can share a VCF, if you let me know if you have a preferred way to send the data. The VCF is from a targeted panel and only contains 517 variants, so hence why I was surprised it was taking so long.

[sigven]

Ok, got it. Just tried the apptainer pull command on our local server and that worked ok, so could it be something with your local setup?

Please share your VCF with me through e.g. Dropbox or e-mail (sigven AT uio.no).

Thanks!

best,
Sigve