Online VEP returns empty columns

Question

[Hamid Ghaedi]

Hi there,

I am trying to annotate a vcf file with ~12k variants using the online VEP. I have tried different options and invariably it returns a table ONLY with "uploaded variant", "location" and "consequence" columns as results. No more annotation fields are populated through the process, and the relevant columns are filled by "-".

I have tried running VEP in docker. It seems this is working however it is super slow (maybe because I am not using cache).

vep --species homo_sapiens --assembly GRCh38 --input_file in.vcf --output_file out_vep.txt --tab  --verbose --database --everything --force_overwrite --fork 60 

Any inputs on why online VEP doesn't do the job?

[Ben_Ensembl]

Hi Hamid Ghaedi,

Do you have a sample of your input file that you can share? I'd be interested to see the input data format and included fields to see if I can replicate the behaviour you are seeing.

[Hamid Ghaedi]

Hey Ben

Sure please find the file here. Also, I have tried online VEP with dbSNP rs ids (rs199612727,rs4984995,rs7496640,rs9776165, rs782403748 ), the same result was obtained: a table ONLY with "uploaded variant", "location" and "consequence" columns

[Ben_Ensembl]

Thanks for sharing your input data. However, I've not been able to replicate the behaviour you are observing. I've run the web-based VEP with default settings/options and your whole input VCF file, the first row of your input VCF file as well as the example rsIDs you provided. All returned expected results with correctly populated tables.

The only other thing I can think of is that you have columns hidden in your results table. There is a button in the top-left hand side of the blue header of the results table. It says (Show/hide columns (x hidden)'. You can click here to show/hide the columns in the output table. Maybe you're hiding all of the columns you want to see?

If that's still not the solution, can you please provide the 'Share' link for the job you have run with the missing columns?

[Hamid Ghaedi]

Here is the link for the result from a very recent try: link.

I guess I am familiar with the show/hide column and almost all of them are populated by "-".

[GenoMax]

I can confirm that running a job with these rs ids produces expected result.

[Hamid Ghaedi]

Perfect, I guess that result comes from running VEP with all default configs, right? However, I was trying to retrieve the most severe consequence for each variant (Restrict results: Show most severe consequence per variant). So almost all the time I only obtained a table with minimum information (variant id, location, and consequence). It seems to me setting "Restrict results" to "Show most severe consequence per variant" is incompatible with almost every other option, something like what we can see for locally installed VEP (here). So instead of a warning/error message, VEP is retuning a table with minimum annotation.

[GenoMax]

Can confirm that adding restrict results filter to online VEP job seems to return a result that matches your original. Note: I only tried the 5 example rsID's and they may not have a "severe consequence" just as an aside.

Answer 1

[Ben_Ensembl]

Yes, using this filter in the web-VEP tool will result in the observed behaviour. It is analogous to using the --most_severe flag in the command line VEP tool. As you can see if the docs, this option forces the VEP to output only the most severe consequence per variant and that transcript-specific columns will be left blank.

You likely want to consider using the 'Show one selected consequence per variant' filtering option in the web-based VEP. This is analogous to using the --pick and --pick_order flags in the command line tool: https://www.ensembl.org/info/docs/tools/vep/script/vep_other.html#pick_eg