Open Stikus opened 3 years ago
Could you fix this, please?
Hi @Stikus,
Thanks for bringing this to our attention. I can reproduce the issue, and I'll let you know when we have a fix in place.
Kind Regards, Andrew
Hi! Any news here? How critical is this inconsistency?
Hello, any progress about this problem?
Hi @serge2016 and @Stikus, Unfortunately we do not have a fix in place for this issue yet.
Very pity( Hope to see it soon!
Hello, any news?
I've found closed PR related to this issue: https://github.com/Ensembl/ensembl-vep/pull/1044
We cannot use latest VEP releases in our production due to this problem because without --no-stats
runtime is twice longer, and we don't have time for such long pipeline.
Until the issue is fixed, as a workaround you could try a solution similar to https://github.com/ntm/grexome-TIMC-Secondary/blob/master/3_runVEP.pl This is designed for the N+1 case, where you regularly run VEP on new VCFs comprised mostly of variants that were seen in your older VCFs. Briefly, this script constructs/updates a private cache of VEP CSQs for each "chr:pos:ref:alt". Then when parsing a new VCF, any line whose "chr:pos:ref:alt" was seen before will not go through VEP, and instead the CSQ is retrieved from the private cache. In our workloads this provides great speedups, and having to disable --no_stats hsan't hurt us much.
Hi @Stikus,
We are still working on this I'm afraid, the pull request you refer to was closed because I don't believe it resolved the issue. We will update when a fix is in place. In the meantime, perhaps the workaround suggested by @ntm above can be useful for you.
Cheers, Jamie.
Dear authors! Could you please add sorting to this place? I hope to see this fix very much!
Hi @serge2016
Rather than repeating, can I point you to my answer on ticket 1066 (https://github.com/Ensembl/ensembl-vep/issues/1066) as it also applies to this issue.
Dear colleagues, @aparton @dglemos @nuno-agostinho @jamie-m-a, a year more is gone. I really think this bug is critical! What can I do to help? I have no idea, how the --no_stats
flag influence results in VCF-file...
Hey @serge2016,
We have not advanced further on this issue, but I'll prioritise this. I am sorry for the inconvenience.
Kind regards, Nuno
Thank you, @nuno-agostinho Hope to see updates soon! It is really critical bug for us...
Dear @nuno-agostinho, Can we hope to see the fix in 110?
Hey @Stikus and @serge2016!
I was playing with a quick fix to the issue (https://github.com/Ensembl/ensembl-vep/pull/1427), but it makes VEP with --no_stats
take as much time as a normal VEP run, so I will have to investigate this further.
From what I can gather, these issues happen only when using --no_stats
together with --hgvs
(may be a variant 3-prime shifting issue). I think the inconsistencies reported could be avoided by simply not using the --hgvs
flag, if that's possible.
I am trying to find a good way to fix this, but the release 110 is imminent (so probably not going in for 110). Sorry for the inconvenience.
Best, Nuno
Hello @nuno-agostinho, have you found the fix for this issue?
Hello, @jamie-m-a !!! Could you help with this, please?
Hi @serge2016 from what we've been able to discover, the --no_stats option when used with --hgvs doesn't properly calculate the hgvs in all cases. The only workaround we've found is to recalculate hgvs at the end, but this will negate the speed improvement offered by running with --no_stats.
Unfortunately we don't have capacity in the team at present to fix this issue, so I'd suggest running without --no_stats and look at parallelisation options (such as using the Nextflow VEP pipeline) to improve annotation speed.
Thank you! I understand, but hope you'll find a possibility to fix this.
Describe the issue
We continue digging into differences (https://github.com/Ensembl/ensembl-vep/issues/1031) caused by
--no_stats
workaround from https://github.com/Ensembl/ensembl-vep/issues/818 and found something more interesting. Two files with--no_stats
and without it differs not only withMAX_AF_POPS
filed (this happening independently of this flag) but in several other lines.Moreover - when we use our old VEP (99) - we don't have this problem (here is working version):
Additional information
When we compared files from 99 version of VEP with 104 version, we found that old files are nearly equal with run result without
--no_stats
- that was another evidence that this flag causing problems in new version.System
Full VEP command line
Data files (if applicable)
Our input file (with little truncated header):
Results: VEP 99 (with our without
--no_stats
- they are equal):VEP 104 (with
--no_stats
):VEP 104 (without
--no_stats
):And here are the differences in transcripts:
First two pair are from first line from VCF and here is the different parts -
G/TE|acg/acGGAg
without--no-stats
andG/GR|acg/ggGAGg
with. But this isGAA
insertion, notGAG
- its strange at least.Last pair has differences not only in amino acids and codons, but in
Consequence
,SIFT
,PolyPhen
too.