Fix H3 vaccine strains

[joverlee521]

Details on Slack.

TODOs

General improvements:

• Standardize capitalization of location in strain names (e.g. DistrictofColumnbia -> DistrictOfColumbia)
• vdb/download: Add --prioritized_seqs_file option #176
• download_from_fauna: Add --prioritized_seqs_file option  seasonal-flu#203

Data clean up:

• remove A/Croatia/10136/RV/2023 sequences since they are duplicates of A/Croatia/10136RV/2023
• remove A/DistrictofColumbia/27/2023 sequences and reupload so they have A/DistrictOfColumbia/27/2023 strain names
• rename all test and reference virus names in titers for A/Croatia/10136RV/2023 and A/DistrictOfColumbia/27/2023

Post clean up:

• update seasonal-flu to use accessions to specify sequences of duplicate strain names

[joverlee521]

Standardize capitalization of location in strain names

This has proven to be more complicated than I initially thought...

I wanted to use geo_synonyms.tsv to standardize locations in strain names, where I would do lower case string comparisons against the label column and then standardize on whatever capitalization was used in the label column. However, there are labels for different locations that would be duplicates in lower case string comparisons, e.g. Hanam/South Korea and HaNam/Vietnam would both match hanam.

This duplicate location label is already an issue for geolocation assignments so I thought I would tackle it here as well. I wanted to use the GISAID provided location metadata to make sure the label and the country matched. Then I realized the virus strain name and the sequence strain name are curated separately and the sequence data do not include the GISAID location metadata. I can create a map of the virus strain name to the GISAID EPI ISL that can be used to match the sequence strain name instead of curating them separately. However, this idea just does not work for the strain names in the titer data because they do not have location metadata. Titer data mostly do not include the GISAID EPI ISL so I cannot think of how to reliably match titer strain name to the sequence strain names...

Even if I can tackle all of the above, this will only standardize strain names for new uploads. If geo_synonyms.tsv is updated, it does not update virus records already in fauna. This is further complicated by the fact that we use strain as the index field for viruses and the virus_strain and serum_strain are used to create the index for titer records.

All that is to say this is taking longer than I would like for fixing this specific data issue, so I plan to do some short term fixes before tackling the larger issue of standardizing locations in strain names.

1. Add entries to flu_fix_location_label.tsv and flu_strain_name_fix.tsv to fix these specific data issues. These files are used in both the vdb and tdb uploads, so these fixes would apply across sequence and titer strain names:

   • DistrictofColumbia -> DistrictOfColumbia
   • A/Croatia/10136/RV/2023 -> A/Croatia/10136RV/2023

2. Remove the "bad" strains from the vdb/flu_* tables and re-upload them to fix their strain names.

3. Remove the titer records with "bad" strains from tdb/* tables and re-upload them to fix their strain names.

[joverlee521]

Manually cleaned up sequence and titer data in fauna and running uploads to S3 in seasonal-flu.