Definition of Private Mutation/Deletion in Nextclade CLI ndjson

[ryhisner]

I'm trying to get some clarity on what the definition of a "private mutation" is in Nextclade CLI. I'm trying to document certain private deletions (which is an amazing addition to Nextclade, by the way), but it seems to be counting deletions that are possessed by entire Pango lineages as private deletions.

Just to give an example, only counting sequences with Nextclade qc scores of ≤ 5, I'm getting 9599 sequences categorized as R.1 (B.1.1.316.1), as having ∆A28271, which is a defining mutation of R.1. There have been 12875 R.1 sequences ever according to CovSpectrum, so this probably means just about all R.1 sequences have ∆A28271 listed as a private deletion by Nextclade. There are also 12736 Delta sequences with ∆A28271 listed as a private deletion as well.

Other lineages that have ∆A28271 as a defining mutation but which list a large number of sequences as having ∆A28271 as a private deletion include B.1.619 (601), B.1.214 (239), B.1.629 (58).

The Nextclade documentation describes private mutations like this:

Does this mean that there are no R.1 sequences at all on the reference tree? Is there a way to fix this? And how is it that so many Delta sequences end up registering ∆A28271 as a private deletion? There are zero Alpha sequences with ∆A28271 listed as a private deletion, so Nextclade gets usually gets this deletion exactly right. I'm not sure why it fails on some lineages.

I don't know if it matters, but I use Wuhan-1 as the reference strain when creating an ndjson file in Nextclade CLI.

[ryhisner]

After looking into it a bit more, I think the problem is that ∆A28271 isn't listed as being a defining mutation of R.1, B.1.619, B.1.214, or B.1.629. Is there a way this could be corrected?

I still have no idea why ∆A28271 appears so often as a private mutation in Delta sequences.

[ivan-aksamentov]

Hey Ryan, looks like everyone is busy at the moment and I am not very well familiar with these lineages to give a useful answer.

I'll ping @corneliusroemer in case he's got a minute.