Would a string composed of `#CHROM`, `POS`, `REF` and `ALT` uniquely identify a variant?

Question

[gernophil]

Hey everyone,

It's a simple question, but the answer might be tricky :) (or not). In my current workflow, I made an unique identifier for vaiants from a VCF composed of these substrings: #CHROM + "_" + POS + "_" + ALT.

For my data this is unique for every variant. But this might not always be the case, would it? There could be a deletion of the G in a GT leading to a T as ALT, but there could also be a simple mutation of that G to a T leading to the same combination of #CHROM, POS and ALT. Or would this change the POS?

So I wanted to extend this by REF to #CHROM + "_" + POS + "_" + REF + "_" + ALT.

So two questions:

1. Is my assumption correct that #CHROM + "_" + POS + "_" + ALT might not always be unique?
2. Would #CHROM + "_" + POS + "_" + REF + "_" + ALT lead to a unique string in every possible case (according to the VCF definitions)?

[LauferVA]

if both the build and the nomenclature system are promulgated

Answer 1

[Pierre Lindenbaum]

well it should be ok in most cases but

• the name of the chromosome might change 1 vs chr1
• you should add the build grch37:1:1234:A:T
• there's symbolic alleles : chr1:77:A:<DELETION>
• multiallelic variants: chr1:777:A:T,C
• normalization : chr1:10:A:ATT vs chr1:10:AA:AAT

[gernophil]

Thanks for the reply :).

• the name of the chromosome might change 1 vs chr1

True, but this will not happen within one analysis, where you mostly stick to the same genome/annotation

• you should add the build grch37:1:1234:A:T

The argument from the above point also counts for here I think. For cross genome uniqueness that would be a good idea.

• there's symbolic alleles : chr1:77:A:<DELETION>

Does this happen often. Isn't a deletion normally shown by fewer ALT then REF bases?

• multiallelic variants: chr1:777:A:T,C

Also true, but I almost always split them.

• normalization : chr1:10:A:ATT vs chr1:10:AA:AAT

This point I don't get. Could you elaborate this?