Clade X pango lineage crosswalk

[tomasleon]

Hi all,

Is there a handy dynamic clade X pango lineage crosswalk file that exists somewhere? We are doing our variant nowcasting by pango lineage and would have to re-classify into clades to submit to this hub. I am hoping this exists but can't seem to find it.

Thanks,
Tomas (CADPH)

[tomasleon]

@nickreich @elray1 not sure how visible discussions are to folks but hoping maybe you can triage to someone who could answer this?

[nickreich]

We are looking into this. Thanks for the ping.

[nickreich]

Hi @tomasleon -
We talked about this as a group today (including some folks from CFA) and here were our thoughts:

1. We don't really think that there is a good, clean way to map pango lineages to clades. This is inherently a noisy, not 1-to-1 mapping. Doing some post-processing to an existing model that operates on pango lineages is possible but likely would be a source of additional error/variation in the model that might be undesirable. That said, changing an existing modeling pipeline may be even less desirable.
2. We note that in the large NextStrain-curated metadata files that we use to establish the "ground truth" for the variant hub there exist, for each sequence, a pango lineage column and a nextstrain clade column. We could imagine that some analysis/cross-tabulation of these data could provide, at the current moment in time, a picture of which lineages map to which clades. There wasn't a sense among our group that this would be a process that we would recommend, but it does seem to be a possible pathway towards seeing how these different lineage/clade "calls" overlap.
3. I've reached out to the Bedford Lab that works on the NextStrain data, as they have a way of mapping clades to lineages (see, e.g. their global sars-cov-2 variant forecast page), maybe they have a method that could be adapted for your purposes.

[rneher]

there is definitely a well defined way to collapse pango-lineages to clades, at least for all clades since 2022. In this tree

https://next.nextstrain.org/nextclade/sars-cov-2

each node is assigned both a Pango lineage and a nextstrain clade and this tree contains (nearly) all Pango lineages. The results should be a one-to-many mapping from Nextstrain clade to Pango lineage.

[nickreich]

Very cool, and I am happy to be corrected by someone more expert in this than I am!

@rneher Do you know of a "simple" downloadable resource similar to what @tomasleon asked for at the top of this thread that provides the "crosswalk" between lineages and clades?

[trvrb]

If you go to https://nextstrain.org/nextclade/sars-cov-2 and click on "Download Data" at the bottom of the page and then click on "Metadata (TSV)" you'll get a TSV file that has columns for both Pango lineage (Nextclade_pango) and Nextstrain clade (clade_nextstrain).

[nlinton]

Thanks, @trvrb. Is there a static URL we can use to download the metadata (tsv) file via code vs having to point and click?

[trvrb]

This particular metadata TSV is generated on-the-fly by the browser and so there's not a simple URL endpoint to get this file. However, if you want a programmatic strategy you can do the following:

nextstrain remote download https://nextstrain.org/nextclade/sars-cov-2/

This will download the JSON file used by Auspice to display tree, etc... as nextclade_sars-cov-2.json. This JSON file contains a tree with tips and each tip will have a Nextclade_pango and a clade_nextstrain attribute. It will be a little bit of work to parse the file, but hopefully not too bad. A full description of this file schema is available here: https://github.com/nextstrain/augur/blob/master/augur/data/schema-export-v2.json, but you might find it easiest to run

nextstrain remote download https://nextstrain.org/ncov/open/reference

and just look at the file.

And alternatively you could work from tree.json here: https://github.com/nextstrain/nextclade_data/tree/master/data/nextstrain/sars-cov-2/wuhan-hu-1/orfs. This is only for the specific-Nextclade dataset, while the above nextstrain remote download solution would work for any page on the site.

[afmagee42]

Hi @rneher and @trvrb, with respect to lining up Pango lineages and Nextstrain clades, I'm having a little bit of trouble squaring what I see in the Auspice tree and in the metadata tsv and hoping you can help me figure out what I'm missing and what the implications are for translating to clades.

The Auspice tree, as promised, appears to neatly and tidily link a given Pango lineage to a single Nextstrain clade. However, in the assigned clades/lineages in the metadata, the correspondence is not so clean.

Looking at the current metadata, sequences labeled BA.2 (full stop, no sublineages) show up with up with 18 Clade labels (plus recombinant), though most BA.2 are 21L as expected over the 21L wave. EG.5.1, KP.3.1.1, and BA.2.86 show up with 6, 5, and 3 (respectively) Pango lineage assignments, though also mostly the expected ones. However, in an older metadata file from May, sequences labeled BA.2.86 are mostly labeled 24A, not 23I.

Is the interpretation here that most of these differences are some form of assignment error? (It is after all an estimate and no estimator is perfect.) Is this some artifact of using metadata from today to look farther back? (I suppose that is perhaps another example of assignment issues, given the warnings about trusting assignments too far in the past?) Something else?

[afmagee42]

Separately, it seems worth considering the potential wrinkle caused by the order of operations between aggregating and translating/mapping. One could conceivably aggregate Pango lineages across an otherwise clean Pango lineage -> Nextstrain clade mapping.

This strikes me as most likely to happen with the "other" grab-bag taxon that shows up in a number of nowcasts, usually based on frequency threshold cutoffs. The "other" obtained by first aggregating Pango lineages and then mapping to Nextstrain clades would seem to me not to be the same "other" that would be obtained directly from clades. Differences are possible in both the composition and total size. Rare (aggregated) Pango lineages are not necessarily rare Nextstrain clades, so unless one sets up a means to break the Pango-based "other" up into its constituents, I would expect Pango lineages to end up in "other" that would otherwise have been assigned to a Nextstrain clade.