Pathways Knowledge Graph

Biological pathways knowledge graph built on Samyama Graph — Reactome, STRING, Gene Ontology, WikiPathways, UniProt

Click for full demo (2:06) — Dashboard, Cypher Queries, and Graph Simulation

Graph Stats

Label	Count	Source
GOTerm	51,897	Gene Ontology
Protein	37,990	Reactome, STRING, UniProt
Complex	15,963	Reactome
Reaction	9,988	Reactome
Pathway	2,848	Reactome, WikiPathways
Total	118,686 nodes

Edge Type	Count
ANNOTATED_WITH	265,492
INTERACTS_WITH	227,818
PARTICIPATES_IN	140,153
CATALYZES	121,365
IS_A	58,799
COMPONENT_OF	8,186
PART_OF	7,122
REGULATES	2,986
CHILD_OF	2,864
Total	834,785 edges

Data Sources

All open-license, human-only (organism 9606):

Source	License	Content
Reactome	CC BY 4.0	2,848 pathways, 9,988 reactions, protein complexes
STRING v12.0	CC BY 4.0	227K high-confidence protein-protein interactions
Gene Ontology	OBO	51K GO terms with IS_A/PART_OF/REGULATES hierarchy
WikiPathways	CC0	Community-curated pathway annotations
UniProt	CC BY 4.0	Protein metadata, gene mappings, disease/drug associations

Quick Start

Option 1: Load pre-built snapshot

# Download snapshot from release
curl -LO https://github.com/samyama-ai/samyama-graph/releases/download/kg-snapshots-v3/pathways.sgsnap

# Start Samyama Graph (v0.6.1+)
./target/release/samyama

# Create tenant and import
curl -X POST http://localhost:8080/api/tenants \
  -H 'Content-Type: application/json' \
  -d '{"id":"pathways","name":"Biological Pathways KG"}'

curl -X POST http://localhost:8080/api/tenants/pathways/snapshot/import \
  -F "file=@pathways.sgsnap"

Option 2: Build from source data

# Install
pip install -e .

# Download all data sources (~1.9 GB)
python -m etl.download_data --data-dir data

# Load into Samyama (all 5 phases)
python -m etl.loader --data-dir data --url http://localhost:8080

# Or selectively
python -m etl.loader --data-dir data --phases reactome string go

Example Queries

-- Top pathways by protein count
MATCH (prot:Protein)-[:PARTICIPATES_IN]->(pw:Pathway)
RETURN pw.name AS pathway, count(prot) AS proteins
ORDER BY proteins DESC LIMIT 10

-- PPI hub proteins (most interaction partners)
MATCH (p:Protein)-[:INTERACTS_WITH]-(other:Protein)
RETURN p.name AS protein, count(DISTINCT other) AS partners
ORDER BY partners DESC LIMIT 10

-- TP53 two-hop neighborhood
MATCH (tp53:Protein {name: 'TP53'})-[:INTERACTS_WITH]-(hop1:Protein)-[:INTERACTS_WITH]-(hop2:Protein)
WHERE hop2 <> tp53
RETURN DISTINCT hop2.name AS protein LIMIT 15

-- Pathway crosstalk (shared proteins)
MATCH (p1:Pathway)<-[:PARTICIPATES_IN]-(prot:Protein)-[:PARTICIPATES_IN]->(p2:Pathway)
WHERE p1.name < p2.name
WITH p1, p2, count(prot) AS shared WHERE shared >= 100
RETURN p1.name, p2.name, shared ORDER BY shared DESC LIMIT 10

-- Immune system sub-pathways
MATCH (child:Pathway)-[:CHILD_OF]->(parent:Pathway {name: 'Immune System'})
RETURN child.name AS sub_pathway ORDER BY child.name

MCP Server (AI Agent Integration)

# Auto-generate MCP tools from the pathways schema
python -m mcp_server.server --url http://localhost:8080

# 12 domain-specific tools: pathway_members, interaction_partners,
# upstream_regulators, drug_pathway_impact, disease_pathways, etc.

ETL Pipeline

Five phases, ordered by dependency:

1. Reactome Core — Pathways, proteins, reactions, complexes, hierarchy
2. STRING Interactions — High-confidence PPI network (score >= 700)
3. Gene Ontology — GO terms, IS_A/PART_OF/REGULATES hierarchy, annotations
4. WikiPathways — Community-curated pathways (deduplicated vs Reactome)
5. UniProt Enrichment — Gene mappings, disease/drug associations

Tests

pip install -e ".[dev]"
pytest tests/

License

Apache License 2.0