Local seqr - Missing HGSV (cDNA and protein) from variants

[gavinmonahan]

Describe the bug
There are no HGSV annotations for cDNA or protein for any of the variants in our local instance of seqr, using the latest containers (also happened in previous versions too). I assume this is an issue with VEP. The only change I made to the containers is updating Hail to 0.2.77 to avoid issue #2460. Is there a way to check the output of VEP/pipeline-runner?
Also, is there a way to add in custom annotations to seqr, ie, an extra field in the INFO column?

Link to page(s) where bug is occurring
NA

Scope of the bug
It's happened for all projects we have made (and all iterations of pipeline-runner).

Screenshots

[hanars]

I agree that this is most likely an issue with VEP, but I would need to know which version of the docker image you are running and which command(s) you used to load the data. The latest docker image includes the correct version of hail, which means you are probably using an outdated version. In terms of checking the output, the pipeline should output an intermediate matrix table in the same directory as the input vcf with the VEP annotations. Checking if that MT has the correct VEP annotations would be a good place to start. I would recommend you check both the vep field in the matrix table and the sortedTranscriptConsequences field

[hanars]

I would ask that you start by updating to the latest version of the pipeline, and make sure you have correctly downloaded all of the necessary reference data. Run docker-compose pull pipeline-runner to get the latest version, and then make sure you follow the initial set up instructions for loading: https://github.com/broadinstitute/seqr/blob/master/deploy/LOCAL_INSTALL.md#annotating-and-loading-vcf-callsets (in particular step 4 of Option 1 or step 3 of Option 2)

[gavinmonahan]

Update: I pulled the latest version and redownloaded the reference data, I had no issue with the pipeline but there are still no HGVScs or HGVSps. I checked the mt for those annotations and there were all "NA". I'm thinking it could be an issue with my VCFs (Ensembl/ensembl-vep#632), but these are all generated and genotyped using GATKs best practices.
I can try running it through VEP 99 then 106 and see if they report the HGVSs. Let me know if you have any advice

Thanks for your help

[hanars]

Thanks for looking into this. VEP 99 is what comes packaged with our pipeline runner, so if you could run it through that and see if that works or not that would be helpful to determine whether there is an issue with the configuration we provide for VEP, or with that version of VEP itself. Sorry for the trouble you are having here and thanks for working with us to solve it!

[gavinmonahan]

Hi @hanars, sorry for the late reply.
I tested my VCFs twice on VEP 104 using the flags --hgvs or --everything, the problem is --hgvs results in exit code 1 if there is no fasta file but --everything simply disables the hgvs annotation steps. I checked the file /vep_configs/vep-GRCh38-loftee.json in the pipeline-runner container and it uses --everything but there was no --fasta flag so I added one:
"--fasta", "/input_vcfs/Homo_sapiens_assembly38.fasta",
After copying the reference genome to input_vcfs.
It resolved the issue and I now have HGVS annotations in seqr :).
Could you please mend this in the container for future uses? I noticed there was another discussion with the same issue as well (#2241)
Thanks for your help!

[hanars]

The latest version of the docker image fixes this issue. Thanks so much for figuring this out and letting me know!

[hanars]

Also, is there a way to add in custom annotations to seqr, ie, an extra field in the INFO column?

No, this is not supported in seqr

[gavinmonahan]

I ask because I'm interested in adding in LOD scores for big families, although I could limit the search to the linkage regions separately it would be convienient to have them on seqr so I can just filter out low LODs

[anjalijain22]

Hi @hanars! I can view the hgvs annotation for grch38 dataset but am unable to view them for Grch37 datasets. I'm using this image of the pipeline runner docker image tagged as latest. The image's digest is: sha256:f2604f5935b2845e5e0fb51b22e1f150f1d128e6bc46af52fb0407d6599e87a6

Just thinking out loud here - do you think a change similar to what @gavinmonahan suggested for /vep_configs/vep-GRCh38-loftee.json would be required for the grch37 config as well?
Would appreciate your help with this!

[hanars]

please feel free to modify your config locally to see if it fixes the issue and then let me know!

[anjalijain22]

Hi @hanars,
I made changes to the VEP config locally and it works, I now have the hgvs annotations!

Similar to Gavin, I -

• Downloaded the GRCh37 reference genome to /mnt/data/seqr/data/vep_data/homo_sapiens directory.
• Unzipped the Homo_sapiens.GRCh37.dna.primary_assembly.fa.gz file
• Modified the /mnt/data/seqr/deploy/docker/pipeline-runner/vep_configs/vep-GRCh37-loftee.json file to add the fasta flag
  "--fasta", "/vep_data/homo_sapiens/Homo_sapiens.GRCh37.dna.primary_assembly.fa"

Would you be able to make this fix in the pipeline runner docker image?

[hanars]

an updated docker image has been pushed up

[michael-cultrera]

Hi @hanars , just wanted to ask a question here as our hgvs annotations are missing. I looked into the matrix table and there doesn't seem to exist a vep field or a sortedTranscriptConsequences field. The --fasta flag exists in the vep config file so not too sure why annotations aren't showing. We have migrated to the new hail backend so not sure if this could be having an impact. Any help would be much appreciated ! Thanks.

[michael-cultrera]

Hi @bpblanken ,

Thanks for for getting back. I noticed that our VEP json config file /var/seqr/vep-reference-data/GRCh38/vep-GRCh38.json references an /opt directory that is empty. For example --fasta /opt/vep/.vep/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz, this file does not exist here, rather it exists in the same directory as the VEP json itself. Going to change this and see if it works.

[bpblanken]

Hi @michael-cultrera,

The '/opt/vep' directory is expected to be empty outside of the docker container. That's the location VEP excepts when VEP runs inside the container. It's created by volume mapping /var/seqr/vep-reference-data/GRCh38/ to /opt/vep/.vep/.

[michael-cultrera]

Hey @bpblanken, makes sense thanks for clarifying. As for your previous comment regarding VEP being misconfigured:

1. VEP reference data seems to look fine inside /var/seqr/vep-reference-data, with there being a vep-GRCh3{7/8}.json within their respective reference genome folders and the json including the --fasta flags.
2. I kept running into JVM related issues when trying to run the hail command you suggested.
3. I got logs that the struct did not have certain fields such as

2025-05-09 19:25:19.712 Hail: WARN: struct{amino_acids: str, biotype: str, canonical: int32, codons: str, consequence_terms: array<str>, exon: str, gene_id: str, hgvsc: str, hgvsp: str, intron: str, mane_select: str, mane_plus_clinical: str, transcript_id: str, lof: str, lof_flags: str, lof_filter: str, lof_info: str, existing_inframe_oorfs: int32, existing_outofframe_oorfs: int32, existing_uorfs: int32, fiveutr_consequence: str, fiveutr_annotation: dict<str, struct{type: str, KozakContext: str, KozakStrength: str, DistanceToCDS: str, CapDistanceToStart: str, DistanceToStop: str, Evidence: str, AltStop: str, AltStopDistanceToCDS: str, FrameWithCDS: str, StartDistanceToCDS: str, newSTOPDistanceToCDS: str, alt_type: str, alt_type_length: str, ref_StartDistanceToCDS: str, ref_type: str, ref_type_length: str}>, am_pathogenicity: float32, spliceregion: array<str>} has no field uniprot_isoform at <root>.transcript_consequences[element] for value JArray(List(JString(Q96NU1-3)))

2025-05-09 19:25:20.068 Hail: WARN: struct{amino_acids: str, biotype: str, canonical: int32, codons: str, consequence_terms: array<str>, exon: str, gene_id: str, hgvsc: str, hgvsp: str, intron: str, mane_select: str, mane_plus_clinical: str, transcript_id: str, lof: str, lof_flags: str, lof_filter: str, lof_info: str, existing_inframe_oorfs: int32, existing_outofframe_oorfs: int32, existing_uorfs: int32, fiveutr_consequence: str, fiveutr_annotation: dict<str, struct{type: str, KozakContext: str, KozakStrength: str, DistanceToCDS: str, CapDistanceToStart: str, DistanceToStop: str, Evidence: str, AltStop: str, AltStopDistanceToCDS: str, FrameWithCDS: str, StartDistanceToCDS: str, newSTOPDistanceToCDS: str, alt_type: str, alt_type_length: str, ref_StartDistanceToCDS: str, ref_type: str, ref_type_length: str}>, am_pathogenicity: float32, spliceregion: array<str>} has no field am_class at <root>.transcript_consequences[element] for value JString(likely_benign)

[bpblanken]

Hi @michael-cultrera,

2. Do you have any logs for the JVM issues you encountered?
3. I'm almost positive those are just warning logs that can be ignored and are safely handled by the seqr implementation of VEP, which drops many fields. If you're able to provide a few specific variants where you're seeing those logs I can try to help debug further!

[michael-cultrera]

Hi @bpblanken,

2. Tried with multiple versions of JVM, got similar output
   A fatal error has been detected by the Java Runtime Environment: SIGSEGV (0xb) at pc=0x000000010b2ac47f, pid=36594, tid=76807 JRE version: OpenJDK Runtime Environment Temurin-17.0.15+6 (17.0.15+6) (build 17.0.15+6) Java VM: OpenJDK 64-Bit Server VM Temurin-17.0.15+6 (17.0.15+6, mixed mode, tiered, compressed oops, compressed class ptrs, g1 gc, bsd-amd64) Problematic frame: V [libjvm.dylib+0xa8647f] Unsafe_GetInt(JNIEnv_*, _jobject*, _jobject*, long)+0xcf No core dump will be written. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again File "/Library/Frameworks/Python.framework/Versions/3.11/lib/python3.11/http/client.py", line 287, in _read_status raise RemoteDisconnected("Remote end closed connection without" http.client.RemoteDisconnected: Remote end closed connection without response
3. Before going ahead, I just wanted to note that our directory that contains the input vcf /mnt/seqr-data/seqr-loading-temp does not include the intermediate matrix table (.mt folder). Hana had mentioned that this mt should be output in the same directory as the input vcf, but I'm not seeing it in that location.