XDK: JN.1.1.1/XBB recombinant(11 seqs, 3 countries)

[xz-keg]

From sars-cov-2-variants/lineage-proposals#1126
JN.1.1.1+A405G, A897C(rev), A1762C(rev), T3431G(rev), C3565T(rev), C5314T

Orf1a: K47R, D211A(rev), L499F(rev), L1056V(rev)

XBB*--JN.1.1.1
Breakpoint between 5315 and 6182

A405G, A897C(rev), A1762C(rev), T3431G(rev), C3565T(rev), C5314T from XBB*

There are many branches of XBB* with 5314T, unable to specify.

GISAID query: A405G, A6183G, C5314T
No. of seqs: 11(France 9 Ireland 1 Belgium 1)

EPI_ISL_18541014, 2023-11-13, France
EPI_ISL_18636383, 2023-12-10, Belgium

usher

[corneliusroemer]

Probably XBB.1.16 as that's the most common XBB with 5341T in Europe by a large margin (>80%)

Could of course have been independently acquired in JN.1.1.1 - but it's true that seems less likely than being from the XBB side.

Actually likely this branch of XBB.1.16.11, common in Europe and also France when this recombination even must have happened:

[FedeGueli]

Designated XDK via 6e78c60

[liamxg]

Dear @aviczhl2,
Could I know which software do you use for detecting recombinations?

[xz-keg]

Dear @aviczhl2, Could I know which software do you use for detecting recombinations?

Currently manual search for reversions on usher https://genome.ucsc.edu/cgi-bin/hgPhyloPlace

I'm developing some code to automatically detect recombs on full usher tree. But don't expect fast as I have other things to do.

I think @JosetteSchoenma may have other ways.

[JosetteSchoenma]

Yes, @liamxg . I use some specific queries to find recombinations with usually one breaking point. A query that finds samples that do not have a specific BA.2.86 mutation at the beginning, but do have one at the end. And another one for the other way around.

Later on, I adjusted those a little, to exclude specific sublineages etc.
After that, I check them in Nextclade for reversions etc. Nextclade adding the 23I clade has been helpful! If the main sublineage is not called BA.2.86* but it has some private mutations with 23I grouped, or the other way around. It is likely a recombinant. If they are scattered, more likely a coinfection or contamination. Usually the latter, certainly from certain countries/labs.

If you look In issue #991 you can see which ones I use specifically.

Hoping to have some time to update their results today.

[liamxg]

Dear @JosetteSchoenma,

Thanks for your reply. hope there are gold reference sequences to be as background, I guess may be a tool for this special problems.

[JosetteSchoenma]

Dear @JosetteSchoenma,

Thanks for your reply. hope there are gold reference sequences to be as background, I guess may be a tool for this special problems.

@liamxg Could you elaborate, please? I do not really understand what you are saying.

[liamxg]

Dear @JosetteSchoenma,

I mean some more sensitive tools to detect recombinants intra-VOCs or intra-VOIs or between VOCs and VOIs.

[JosetteSchoenma]

Dear @JosetteSchoenma,

I mean some more sensitive tools to detect recombinants intra-VOCs or intra-VOIs or between VOCs and VOIs.

What about them?

[yatisht]

I was wondering if you are aware of RIVET.

Manuscript: https://academic.oup.com/bioinformatics/article/39/9/btad538/7257069
Website: https://rivet.ucsd.edu/

Below is the SNP plot for the closest match to your description that I found on our latest RIVET analysis.

The corresponding node id corresponding to this node on the Dec 17 UShER-MAT is node_1453233. It is inferred as a recombinant between JN.1.1.1 with XBB.1.18.1 or XBB.1.28.1 or XBB.1.9.1. The predicted breakpoint interval is (3565,7842) or (3565,6183) or (3565,5720) for the three possible pairs, respectively. The predicted date of origin using Chronumental is 2023-11-12.

I have also attached the list of descendants (node_1453233_descendants.txt).

Also, we would welcome feedback or suggestions for RIVET to aid your analysis.

[xz-keg]

Also, we would welcome feedback or suggestions for RIVET to aid your analysis.

It seems only use designated lineages as reference instead of all lineages on usher? Usually recombs have to be sub-lineage accurate to achieve the fewest private mutations possible. (Here using XBB.1.16.11+C5314T branch gives 0 private mutations while using designated lineages give at least 1 private mutation).

[yatisht]

Hi @aviczhl2, we do indeed use all sequences in the MAT (including ancestral corresponding to the internal nodes of the UShER-MAT) for the recombination analysis. RIVET uses the RIPPLES algorithm (https://www.nature.com/articles/s41586-022-05189-9) for searching exhaustively. The acceptor and donor sequences in RIVET correspond to some internal nodes in the UShER-MAT. Does that answer your question?

[FedeGueli]

thanks @yatisht for the suggestion and explanation

[xz-keg]

Hi @aviczhl2, we do indeed use all sequences in the MAT (including ancestral corresponding to the internal nodes of the UShER-MAT) for the recombination analysis. RIVET uses the RIPPLES algorithm (https://www.nature.com/articles/s41586-022-05189-9) for searching exhaustively. The acceptor and donor sequences in RIVET correspond to some internal nodes in the UShER-MAT. Does that answer your question?

Oh, I see. You seem to be using the node before C5314T as reference?

[yatisht]

That is right. Please see the subtree below:

RIVET is picking the leftmost node (node_1453233) carrying A405G, A897C, A1762C, and T3431G as the recombinant node. Looks like C5314T mutation happens in its descendant and is not shared by all descendants.

UShER subtree: link

[xz-keg]

That is right. Please see the subtree below:

RIVET is picking the leftmost node (node_1453233) carrying A405G, A897C, A1762C, and T3431G as the recombinant node. Looks like C5314T mutation happens in its descendant and is not shared by all descendants.

UShER subtree: link

Why do use this? This is a recombinant island. The top large branch is XDD, middle are numerous different singlet BA.2.86*/XBB* recombinants, only the C5314T branch refers to XDK.

[yatisht]

I spoke to @aviczhl2 today over a call and we concluded that RIVET can indeed struggle to accurately identify recombinants in a complicated recombinant island such as this example. However, it can still be useful to locate these recombinant islands. We will try to add some features to flag suspicious recombinant islands.

[liamxg]

Dear @yatisht,
cool! any example for this, I have projects need analysis recombinants.

[corneliusroemer]

@yatisht To filter out using spurious branches/islands or artefacts as candidate donors and be more robust, you could try using the Nextclade reference tree composed of only Pango lineages that have been checked to be real. Of course you loose sensitivity, but gain in specificity. Might also make it run much faster. I don't know what the input is that RIVET expects, if it's a MAT one could simply turn the Auspice json into a MAT (json at wget --output-document tree.json "https://nextstrain.org/charon/getDataset?prefix=staging/nextclade/sars-cov-2").

[liamxg]

Dear @corneliusroemer,
MAT = mutation-annotated tree, right.