Geographic fixes

[jameshadfield]

Please update this issue as updates are made / we notice things

Updated after #39, but the changes here aren't just because of that, geographic info on the PPX side has also improved

---

There's a few aspects to this:

• [ingest] Adding missing (geographic) metadata via annotations_geo.tsv
• [ingest] Fixing typos / formatting via geolocation_rules.tsv
• [ingest] Fixing incorrect metadata
• [phylo] Adding missing lat/longs. These are shown as warnings when running the export rule

Missing geo metadata

Ideally we'd get complete country & division for all samples. Within ingest/ running ./scripts/summarise_geography.py --m1 data/metadata_geo_improvements.tsv will summarise the geographic combinations present which makes it easy(ish) to see missing countries, divisions etc. Quite a few of the missing values are from sequence fragments, so filtering to 18kb (as we do in the phylo workflows) makes our task easier.

augur filter --sequences results/sequences.fasta --metadata data/metadata_geo_improvements.tsv --metadata-id-columns accession --min-length 18000 --output-metadata data/check-geo.tsv
./scripts/summarise_geography.py --m1 data/check-geo.tsv

Here's a summary of that output focusing on combinations which are missing country/ division, sorted by count: (no strains are missing country any more)

num	country	division	location
775	Democratic Republic of the Congo
454	Sierra Leone
99	Gabon
16	Liberia
10	Guinea
4	Uganda
2	Mali

Incorrect geo metadata

(From #24): The 2 Kelle samples (PP_000LEP0, PP_000LEQY) have country of DRC but Kelle is in Republic of Congo. This needs investigating.

[joverlee521]

(From #24): The 2 Kelle samples (PP_000LEP0, PP_000LEQY) have country of DRC but Kelle is in Republic of Congo. This needs investigating.

This has also been reported in pathoplexus/curation_reports#8