NeuroGhost

A shared vocabulary for neuroscience data.

---

NeuroGhost is a public catalog of neuroscience vocabularies. Labs publish their LinkML schema; the registry compares it to every other schema and surfaces which terms mean the same thing across projects.

Distance score — 0.0 = identical, 1.0 = unrelated. Computed via the Proteus pipeline: name similarity, token Jaccard, alias overlap, definition embeddings, IRI anchor, and unit dimensional veto. Adjustable live on the Concepts page.

---

Website

sensein.group/NeuroGhost — seven tabs: Concepts, Diff, Graph Schema, Transform, Query, Provenance, Register. Every view has download buttons.

---

API

Static JSON via GitHub Pages — no auth, no rate limits, CORS open.

Method	URL	Status
GET	/data/registry.json	✅ Live
GET	/data/versions/{version}.json	✅ Live
GET	/data/provenance.json	✅ Live
GET	/api/transform?from={schema}&to={schema}	🔜 Planned
POST	/api/transform	🔜 Planned

distance: 0.0 = identical · 1.0 = unrelated.

Response shapes

GET /data/registry.json

{
  "registry_version": "1.7.0",
  "generated_at": "2026-07-23T12:40:24Z",
  "sources": [{ "label": "bbqs", "version": "1.0.0", "class_count": 29 }],
  "classes": [{
    "hash_id": "sha256:abc123...",
    "iri": "https://registry.sensein.io/obj/Subject",
    "name": "Subject",
    "definition": "A research participant.",
    "sources": ["bbqs"],
    "properties": [{ "hash_id": "sha256:def456...", "name": "age", "value_range": "xsd:integer" }],
    "alignments": [{ "target_name": "Participant", "distance": 0.12, "method": "composite" }]
  }]
}

GET /api/transform?from=bbqs&to=bids (planned)

{
  "from": "bbqs", "to": "bids",
  "mappings": [{
    "from_class": "Subject", "to_class": "Participant", "distance": 0.12,
    "field_mappings": [
      { "from_field": "subject_id", "to_field": "participant_id", "confidence": 0.85 }
    ]
  }]
}

POST /api/transform (planned — needs serverless layer)

curl -X POST https://sensein.group/NeuroGhost/api/transform \
  -H "Content-Type: application/json" \
  -d '{ "from": "bbqs", "to": "bids", "data": { "subject_id": "sub-01", "age": 24 } }'

---

How alignment works

Alignment runs the Proteus pipeline (github.com/neurovium/Proteus) — inlined into neuro_ghost/align.py. Six stages:

Stage	Name	What it does
0	Load	Reads every class from LadybugDB into a MatchingProfile (name, aliases, IRI, units, definition)
1	Block + Unit veto	Generates candidate pairs across schema pairs (recall-focused). Hard-vetoes pairs whose units have known but incompatible SI dimensions (e.g. Hz vs V). This is the only precision filter at this stage.
2	SignalVector	For each candidate pair, computes a frozen evidence bundle: name similarity, token Jaccard, alias overlap, definition cosine (sentence-transformers), unit compatibility, anchor relation (IRI match). Absent signals are None, never 0.0.
3	Calibrate	Weights the present signals into a confidence score. Weights: name 0.45, token Jaccard 0.35, alias overlap 0.20 (renormalized when signals are missing). Adds a 0.05 bonus for known-compatible units. Blends in definition similarity at 25% when embeddings are available.
4	Predicate	Two-pathway assignment. Anchored (IRI evidence): can reach skos:exactMatch, skos:broadMatch, skos:narrowMatch. Statistical (no IRI anchor): caps at skos:closeMatch. Pairs below 0.45 confidence are dropped.
5	Repair	Structural cleanup — demotes duplicate exactMatch claims to closeMatch. Never deletes edges.
6	Write	Writes ALIGNED_TO edges in LadybugDB with distance, skos_relation, method, and per-signal subscores.

Distance is 1 − confidence, so 0.0 = identical, 1.0 = unrelated.

Definition embeddings use all-MiniLM-L6-v2 and are cached in data/embeddings.parquet so CI doesn't recompute from scratch.

---

Adding a schema

1. Write a LinkML .yml file (copy schemas/bbqs.yml as a template).
2. Go to the Register tab, paste your YAML, click Open GitHub Issue.
3. A GitHub Action validates, ingests, aligns, and archives it within minutes.

No installation, no pull request, no reviewers required.

---

Running locally

git clone https://github.com/sensein/NeuroGhost.git
cd NeuroGhost
pip install -r requirements.txt

python neuro_ghost/pipeline.py --fresh                              # full rebuild
python neuro_ghost/pipeline.py --fresh --skip-converters            # local schemas only
python neuro_ghost/pipeline.py --skip-converters --schemas schemas/bbqs.yml  # one schema

Options: --fresh (wipe DB), --skip-converters (skip BIDS/NWB/DANDI/openMINDS/AIND fetch), --schemas FILE, --bump major|minor|patch, --agent TEXT.

Open index.html in a browser when done.

---

Stack

• LadybugDB — embedded graph DB, no server
• LinkML — schema format
• sentence-transformers — all-MiniLM-L6-v2 for semantic distance
• Static HTML + GitHub Pages — one-file frontend, no framework
• GitHub Actions — CI/CD on every schema submission

---

Contributing

• Register a schema via the Register tab.
• Open an issue to report bugs or suggest features.
• PRs welcome, especially around the distance function.

License: CC0-1.0 — public domain.