Capture uploaded allele correctly for VCF input

[likhitha-surapaneni]

Ticket: ENSVAR-5858

• Fix Unit tests
• Review header and output columns

To test:

• With --allele_number for multi-allelic examples

[nakib103]

Multi-allelic is not getting minimised for default format. For example - 1 961320 961324 GCAGG/GCA/GCAG +

But in the output still getting MINIMISED=1, (without the PR they are also not minimised but there is no MINIMISED=1).

[likhitha-surapaneni]

Hi @nakib103 , can you please test this example with the latest commit. The allele is expected to be similar to when running --minimal

[nakib103]

Hi @likhitha-surapaneni,

Yes, the allele are same when running with or without --minimal.
But the original problem remains. The output says it is minimised but when it is not -

1_961320_GCAGG/GCA/GCAG	1:961320-961324	-	ENSG00000188976	ENST00000327044	Transcript	upstream_gene_variant	-	-UPLOADED_ALLELE=GCAGG/GCA/GCAG;IMPACT=MODIFIER;DISTANCE=2067;STRAND=-1;MINIMISED=1

[likhitha-surapaneni]

Hi @nakib103 , the output is still minimised if you notice the 3rd column (correct me if I got it wrong). We have the allele "-" as the minimised representation for first alternative allele is GG/- and for the second one is G/-. The problem however is that there is no way to differentiate between the alternative alleles as both show "-". Ideally minimal representation should be GG/-/G. This is also an existing problem with the option --minimal and needs to be addressed probably in a future ticket. Please let me know if this makes sense

[nakib103]

I was actually looking at the Uploaded_variation column, it does not seem to be minimised as it does for bi-allelic variants (see - 1 961320 961324 GCAGG/GCA +).

But the Allele column shows the alleles are minimised. It seems Uploaded_variation string has different logic, should we address this?

[likhitha-surapaneni]

Yes, I agree @nakib103. The uploaded variation was not expected to minimise alleles but it was minimising the alleles in some cases due to the way we populate the column and it was difficult to correct this as it is also the default option, hence the approach was taken to capture original allele with a different flag (--uploaded_allele). However, I agree it may also be worth investigating why uploaded variation is minimising the allele in some cases and not doing it in other cases.

[dglemos]

The MINIMISED is missing from the header, so in the VCF output even when the variant is minimised this info is not in the file. The only outputs with this info are the default and json.

[dglemos]

When the input is in VCF format, example:
1 961320 . G GCAGGCTCGGCC . . .

In the vep default output, the Uploaded_variation is represented with minimised alleles:
1_961321_-/CAGGCTCGGCC
In this scenario matching the alleles is more difficult.

[likhitha-surapaneni]

When the input is in VCF format, example: 1 961320 . G GCAGGCTCGGCC . . .

In the vep default output, the Uploaded_variation is represented with minimised alleles: 1_961321_-/CAGGCTCGGCC In this scenario matching the alleles is more difficult.

Hi @dglemos , yes this is right. The idea is to use --uploaded_allele to extract the original allele through UPLOADED_ALLELE=G/GCAGGCTCGGCC (docs). Please let me know if that would be sufficient to match the alleles.

[dglemos]

Yes, UPLOADED_ALLELE is enough to match the alleles. It should be returned by default though, if the variant is minimised it should return the original alleles too.

[nakib103]

Yes, UPLOADED_ALLELE is enough to match the alleles. It should be returned by default though, if the variant is minimised it should return the original alleles too.

I think it is generally advisable to not update the default output list so not to break existing pipelines.