Update usage of "accession" as the ID column

[victorlin]

Description of proposed changes

New options in Augur 22.1.0 allow usage of this column as the ID column across all subcommands that read metadata.

For this workflow in particular, the metadata file can now be used as-is. This removes the need for a modified copy of the metadata. It also allows specifying an original metadata column "strain" as the display strain field, rather than a column "strain_original" generated during the Snakemake workflow.

Related issue(s)

• Proposed in Allow selection of metadata ID column augur#1237

Testing

• Checks pass
• Manually compared Auspice JSON visualized in Auspice. Tip attributes are identical. (ref)

[j23414]

Not sure how to test this properly.

May be able to check this manually, although I'm all for a reasonable automated check!

1. Update to the latest nextstrain version to get the --metadata-id-columns flag

nextstrain update

2. Compare runs, compare a few files and make sure nothing breaks

git clone https://github.com/nextstrain/rsv.git original
git clone https://github.com/nextstrain/rsv.git changed

cd original
nextstrain build .

cd ../changed
git checkout victorlin/use-custom-metadata-id-column
nextstrain build .

3. Mouse over the tips in the tree. Is strain or accession displayed as tips in the two sets of trees? Do they match?

nextstrain view auspice
nextstrain view ../original/auspice

Or compare changed against the live RSV site

• https://next.nextstrain.org/rsv/a/genome

[victorlin]

Thanks @j23414! I'll give it a try.

[victorlin]

I added CI with example data to get a quick build. Looks like accession is missing from the Auspice JSON. I'll investigate more tomorrow.

[j23414]

Oh, thanks for catching! And I like the fixup. For the missing accession, might be able to adjust

rsv/scripts/set_final_strain_name.py

Lines 6 to 7 in 77840b2

	if node["name"] in lookup:
	node["name"] = lookup[node["name"]]

to something like:

 if node["name"] in lookup:
     node[metadata_id_columns] = node["name"]
     node["name"] = lookup[node["name"]] 

I haven't tried the above code so I'm not certain if adding new attributes to a node would cause downstream errors.

[victorlin]

I think it has something to do with rule export which comes before rule final_strain_name, since accession is missing from tree.json as well.

I didn't get around to this yet and might not today, but I'll update here when I do!

[victorlin]

Ok, I think I got to the root of this. Reasons:

1. In augur export, the accession node attribute gets exported as long as it's an available node attribute.
2. Among other sources, node attributes come from metadata.
3. The metadata used above comes from read_metadata, which does not make the index column name accessible like the others.
4. This PR is updating the index column to be accession, making it inaccessible as a node attribute due to all of the above.

If this is it, then a proper fix would be to update read_metadata to keep the index column as a column when setting the index.

Signing off now, I'll get to this another day.

[victorlin]

d123d60's PR-triggered CI run results (pr29/d123d60/docker/rsv/a/F, etc.) should be comparable to those from latest master (master/4ecf498/docker/rsv/a/F, etc.).

[tsibley]

(tanglegram)

[j23414]

Ran a manual test, I can see both strain name and accession named tips. Looks good to me!