[bendichter]

Addresses #1979.

What this does

• pynwb_utils._get_herd_anatomy reads the file's HERD table (nwb.external_resources) and collects every entity with a UBERON: or MBA: ID, whatever object or field it is linked from. Entities linked only from Subject objects are skipped: UBERON also holds life-stage terms that may be linked from Subject.age. The result goes into the metadata dict as herd_anatomy.
• brain_areas.herd_anatomy_to_models turns those into deduplicated Anatomy models. HERD stores no labels, so the name comes from the bundled Allen CCF table for MBA: terms, else the longest value in the file linked to the term (so trapezius muscle beats EMG), else the entity ID.
• _extract_brain_anatomy uses HERD anatomy first, for any species, and falls back to the existing mouse-only string matching from feat: extract brain area anatomy from NWB location fields #1807. The terms land in the same place: wasDerivedFrom[BioSample].anatomy, on the synthetic brain-region-sample when there's no tissue/slice/cell ID.

The HERD tables are read directly rather than through HERD.to_dataframe(), because that call raises a pandas deprecation warning that the test config turns into an error.

Checked on DANDI:001888

On sub-GRIN2B-129/sub-GRIN2B-129_ses-2021-03-26_ecephys.nwb (rat, so #1807 extracts nothing), streamed from S3:

Identifier	Name
UBERON_0008933	S1Hl_S1Fl
UBERON_0016634	M2_anterior
UBERON_0001384	M1_anterior
UBERON_0022232	V2_ML
UBERON_0002436	V1_M
UBERON_0002380	trapezius muscle
UBERON_0000956	cortex

The species, strain and genotype links in the same table are left out. Most names are lab abbreviations. Once #1825 lands, its bundled UBERON table can supply real labels through _ontology_label; I left that out to keep this PR independent of #1825.

Not done here

• The open questions in Extract asset anatomy from HERD ontology links in NWB files #1979: dropping broader terms like cortex when more specific ones are present, and stimulus body parts.
• Aggregating anatomy into AssetsSummary, which needs a dandi-schema change.

Tests

• New tests: a rat file with HERD (anatomy extracted, Subject.age life-stage term skipped), a mouse file where HERD wins over Allen matching, label/dedupe rules, and HERD anatomy without a species.
• pynwb < 4 can't store HERD in the NWB file, so the two file-writing tests skip there; I ran them under pynwb 3.1.0 / hdmf 4.1.0 (skipped, rest pass) and pynwb 4.2.0 (all pass).
• test_metadata.py, test_pynwb_utils.py and test_brain_areas.py pass locally except three BIDS tests that need bids-validator-deno, which isn't installed here (they fail on master too). mypy and pre-commit are clean.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 93.98496% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.26%. Comparing base (54e842c) to head (4f911c6).

Files with missing lines	Patch %	Lines
dandi/pynwb_utils.py	90.24%	4 Missing ⚠️
dandi/metadata/brain_areas.py	89.28%	3 Missing ⚠️
dandi/tests/test_metadata.py	98.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1980      +/-   ##
==========================================
+ Coverage   91.18%   91.26%   +0.08%     
==========================================
  Files          95       95              
  Lines       14740    14872     +132     
==========================================
+ Hits        13440    13573     +133     
+ Misses       1300     1299       -1     

Flag	Coverage Δ
unittests	91.26% <93.98%> (+0.08%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.