The BioBTree knowledge graph

BioBTree already integrates 80+ primary databases into its own high-performance graph database — resolving identifiers and linking cross-references across every source. Rather than re-mining the web, the knowledge-graph export reads BioBTree's pre-built internal index files directly and re-expresses them in the biolink model: each dataset is mapped to biolink categories and typed predicates, edges are annotated with evidence, and the result is written as standard KGX. Because it builds on BioBTree's existing federation, the graph inherits its identifier resolution and regenerates end-to-end whenever the underlying data is refreshed.

1. 1

   BioBTree database

   80+ primary sources already integrated and cross-referenced in BioBTree's pre-built index.

2. 2

   Map to biolink

   Each BioBTree dataset maps to biolink categories and typed predicates via versioned mapping tables.

3. 3

   Extract & link

   Nodes and edges are read straight from BioBTree's index, identifiers resolved to canonical CURIEs.

4. 4

   Add evidence

   Edges carry ECO evidence codes, qualifiers and the contributing source.

5. 5

   Assemble & export

   Merge, dedup, stub and validate, then emit KGX nodes / edges (TSV + JSONL) with a manifest.

Two graphs — the full source, and a graph you can load

// gene → disease (1 hop)
MATCH (g:Gene {name:'BRCA1'})-[r]-(d:Disease) RETURN type(r), d.name;

// drug → target protein → gene (drug pharmacology)
MATCH (drug)-[:affects]->(p:Protein)<-[:has_gene_product]-(g:Gene {name:'EGFR'})
RETURN drug.id, drug.name;

// filter on node attributes
MATCH (n:Gene) WHERE n.gnomad_pli > 0.9 RETURN n.name ORDER BY n.gnomad_pli DESC;