VT Normalize skips multi-allelics

[AlistairNWard]

I'm running through the demo data set and seeing a SNP with a non-canonical representation. vt normalize should not output this variant, so are we sure it's running?

Also, if there is no annotation (e.g. Sift, PolyPhen) should we write 'no annotation', or something? See how everything is blank in the image?

[AlistairNWard]

[tonydisera]

I'm looking into this now. Sad!

[tonydisera]

This is a multi-allelic variant. Doesn't vt normalize ignore these?

I can't find an issue on this problem, so we should probably try to address this in a future release.

[AlistairNWard]

DAMN IT!!!!! I've been caught out by this a number of times. It is an issue, but not one we've prioritized. vt doesn't ignore it, but since the multiallelic obviously contains a deletion as well, this is the shortest representation that is consistent with all alleles. If we have a multiallelic, we should probably, as a short term solution (but not this release), include all alleles at the top. I saw this as a SNP, but then the DNA is nuts. If I saw all alleles, it would hopefully stop me making this mistake!