Pango XM weirdness

[hyanwong]

Most of the Pango X lineages I have looked at are very clean, but XM is weird (pink samples in the graph below). It appears we are claiming that there are 5 separate origins of XM, which all have breakpoints in very similar places. I wonder if one or two of these would disappear if we seeded with e.g. SRR18814882, or if the XM samples had been added in a different order?

[jeromekelleher]

The first XM in the ARG (SRR19024311) went in with zero mutations, so I don't think that's the answer

Here' the XM pango designation issue for reference: cov-lineages/pango-designation#472

[hyanwong]

Here's a more comprehensive subgraph with deletions (apologies for how messy it looks: it's hard to avoid crossing lines in this one).

The long branch of magenta "insertions" (reversion of a deletion) at the bottom left indicates that we should probably have placed this one somewhere that does not have the 21633-21641 deletion (some of the other XMs do not have this, I think)

[hyanwong]

If we look at the copying pattern, it appears that the deletion distribution confirms that at least some of these are indeed separate events (below, recombinants ordered left-to-right). It might be worth digging into this, if it provides confirmation that we are doing the right thing?

[jeromekelleher]

A lot of independent high frequency deletions going on here! Seems pretty weird to me that these would have got to such high frequency (to permit inclusion) if they only happened in this (I assume) relative backwater of the ARG. They must be recurring a lot, then?

[hyanwong]

I'm not sure these deletions are independent: I think they are being inherited from a (slightly distant) ancestor. At least, you can't see many recurrent deletions (black fill, coloured outline) in the subgraph, apart from in the BA1 and BA2 ancestors.

I think the best thing here would be to plot out the deletions in each of the putative independent XM clades (plus the weird long-branch one), and see if they look independent.

[hyanwong]

I also just noticed that the 3 left hand XM origins, and also the one second from the right, do not have the deletion seen in the main clade (21633-21641), because their breakpoints are above 21641, and so are more parsimonious placements for the weird reversion-of-deletion sample at the end of the magenta-studded line.

I.e. if we had included deletions in the matching, we would surely have placed that sample within one of the other XM clades. I guess we could test this easily enough.

[szhan]

Just looking at the CovRecomb results.

It seems like they suggest XM to be of multiple independent origins, based on their criteria, as follows:
(1) They clustered the putative recombinant samples by the parent pairs.
(2) They sorted the samples by their collection dates. The first sample is considered as the initial recombinant event.
(3) Later collected samples are considered independent events if:
(a) they are separated from the previous sample by 30 days or more,
(b) they differ from the previous sample by at least 4 mutations.

All the independent events have left and right parents as BA.1.1* and BA.2.23, respectively.

deletions_on_left	deletions_on_right	country	collection date
6513_3D, 11285_9D	21633_9D	USA	2022/1/29
6513_3D, 11285_9D	21633_9D	Germany	2022/3/7
6513_3D, 11285_9D	21633_9D	Germany	2022/2/22
6513_3D, 11285_9D	21765_6D, 21987_9D, 22194_3D	USA	2022/1/9
6513_3D, 11285_9D	21765_6D, 21987_9D, 22194_3D	France	2022/3/28
6513_3D, 11285_9D	21633_9D	Netherlands	2022/3/23

** xD indicates x bases deleted.

The deletions roughly corresponding to those in Yan's copying patterns are highlighted. We are missing 22194_3D, it seems.

Note that they analysed 14m sequences from GISAID sampled up to Jan., 2023. In our ARG, the Viridian batch 1 samples go up to Feb., 2023.