# Neuro Atlas > An interactive, evidence-graded atlas of human signalling systems: neurotransmitters, hormones, receptors, genes, brain > structures and organs — where each signal is made, where its receptors are expressed, what it connects to, and what the > peer-reviewed literature establishes about it. Educational; not medical advice. Everything is static JSON with stable ids (no key, no rate limit — please cache). Relative paths below resolve against https://neuro-atlas.si/ . Human-readable documentation: https://neuro-atlas.si/#/data ## Core files - [findings.json](https://neuro-atlas.si/data/findings.json): "what's known" — {"version":1, "findings": {: {topic, finding, basis, scope, pmids, grade, species}}}. basis: supported (established) | corrected (the narrower/opposite version the papers support) | mixed | unsettled (not established). scope "out" = not about a biological system; ignore those. grade: high | moderate | low | very-low. - [interactions.json](https://neuro-atlas.si/data/interactions.json): subject–relation–object triples stated by the findings — {"version":1, "edges": [{s:{type,id|label}, p, o:{type,id|label}, context:{species?,condition?}, finding, basis, grade, pmids}]}. Id types: molecule, gene, structure (brain mesh or region-group id), organ, concept, drug_class, metabolite (ChEBI id, with label), complex (multi-subunit receptor/channel/kinase, e.g. nmdar, gabaa, mtorc1; members in complexes.json). Label types: cell_type, condition, behaviour, process, intervention, place. Relations: increases, decreases, activates, inhibits, binds, modulates, is_produced_in, is_expressed_in, projects_to, is_required_for, contributes_to, protects_against, is_associated_with, predicts, is_marker_of, is_part_of, is_disrupted_in, is_converted_to, has_no_effect_on (tested null), is_not_associated_with (observational null). An edge records what a finding states, weighted by its basis and grade — not a curated pathway. - [molecules.json](https://neuro-atlas.si/data/molecules.json): signalling molecules keyed by IUPHAR/BPS GtoPdb ligand id ("gtopdb:L2174" = oxytocin): receptors (GtoPdb targets with HGNC genes, action, affinity, PMIDs), Reactome pathways, precursor gene, and the ids of findings about it ("claims"). - [genes.json](https://neuro-atlas.si/data/genes.json): genes keyed by HGNC symbol: UniProt, Ensembl, Human Protein Atlas mRNA by tissue / brain region (nTPM) / cell type (nCPM), Allen Human Brain Atlas z-scores per HRA structure. - [body.json](https://neuro-atlas.si/data/body.json): organs (atlas ids with Uberon/FMA): which featured signals are made / sensed there (expression-derived), and finding ids. - [guide.json](https://neuro-atlas.si/data/guide.json): everyday functions (sleep, appetite, mood, bonding…) — plain explanation, signals involved with their role, structures and organs; every line cites a PMID or database entry. - [concepts.json](https://neuro-atlas.si/data/concepts.json): phenomena (stress response, pair bonding…) with finding ids and top molecules. - [studies/index.json](https://neuro-atlas.si/data/studies/index.json): the papers findings rest on (compact rows; fields listed in the file). Full records with abstract excerpts and stance (supports | qualifies | contradicts) per finding: data/studies/.json. Retracted papers are flagged and never count as support. - [core.json](https://neuro-atlas.si/data/core.json): the web app's compact column-encoded index (includes brain structures keyed by HRA mesh id with Uberon terms). Prefer the files above. - Bulk: data/open/findings.jsonl.gz and data/open/interactions.jsonl.gz (one in-scope finding / one edge per line). ## Stable ids and page links - Molecule gtopdb:L → https://neuro-atlas.si/#/m/gtopdb:L2174 - Gene HGNC symbol → #/g/OXTR · Brain structure HRA mesh id → #/s/amygdaloid_complex · Organ → #/o/adrenal-glands - Function → #/f/sleep · Phenomenon → #/c/pair-bonding · Paper PMID → #/p/20538951 - Finding: opaque record id (key of findings.json; `finding` field of an edge; listed under "claims" in molecules/body/concepts). ## Using it to cross-reference and hypothesise 1. Pick an entity id (e.g. gene OXTR). 2. Collect edges where s.id or o.id equals it. 3. Join the other endpoints to molecules.json (receptors, pathways) and genes.json (where they are expressed). 4. Keep only edges with basis "supported" (or weight by grade) and check species in `context` — animal and in-vitro evidence is labelled, not human fact. 5. Two-hop paths (A → X → B) that no single finding states are hypotheses to test against the primary literature, not facts. ## Chains ("If this, then…") The atlas computes live cause→effect paths through the interaction graph using stable node keys (`sk`/`ok` fields on each edge, e.g. `molecule:gtopdb:L2174`, `gene:OXTR`, `label:slow wave activity in nrem sleep`). The Connect tool (#/connect) finds paths between any two entities. Sign rules: - Positive (+): increases, activates, contributes_to, is_required_for - Negative (−): decreases, inhibits, protects_against - Mechanistic (0): binds, modulates, is_converted_to, projects_to, is_disrupted_in — direction not asserted - Correlational (~): is_associated_with, predicts, is_marker_of — weaker than causal - Location (excluded from traversal): is_produced_in, is_expressed_in, is_part_of - Null (terminates chain): has_no_effect_on, is_not_associated_with — shown as "tested: no effect" Composed sign: null overrides all; then ~ if any step correlational; then 0 if any step mechanistic; else product of +/− (+ × − = −; − × − = +). Weakest-link strength = minimum grade rank + worst basis along the path. Any chain of ≥ 2 edges is marked "inferred" — the end-to-end effect was not directly tested in a single study. Steps where the species context changes (human → animal:rat) are flagged. Two-hop paths you compute from interactions.json are hypotheses to test, not established facts. ## Licences - Our text (findings, interaction triples, function guide, plain-language lines): CC BY 4.0 — attribute "Neuro Atlas, https://neuro-atlas.si". - BodyParts3D (3D body), © The Database Center for Life Science: CC BY-SA 2.1 JP. HuBMAP Human Reference Atlas (3D brain): CC BY 4.0. - Abstract excerpts are brief quotations from third-party papers (rights remain with the publishers); each carries its PMID. - Reference data keep their own terms: IUPHAR/BPS Guide to PHARMACOLOGY, Human Protein Atlas, Allen Human Brain Atlas, HGNC, Reactome, Uberon. Not medical advice: findings report what the literature shows; they are not dosing or treatment recommendations.