Bionex

Biomedical knowledge graph infrastructure. Integrates 67 datasets into a unified graph and generates network embeddings.

Based on Bioteque (IRB Barcelona, MIT license). Citation: Fernandez-Torras et al., Nature Communications (2022). doi:10.1038/s41467-022-33026-0

Architecture

%%{init:{'theme':'base','themeVariables':{'primaryColor':'#f8f9fa','primaryTextColor':'#1a1a2e','primaryBorderColor':'#adb5bd','lineColor':'#6c757d','fontSize':'13px'}}}%%
graph LR
    subgraph Sources[" 67 External Sources "]
        direction TB
        S1(["DrugBank · STRING · LINCS"])
        S2(["GPSAdb · DisGeNET · CCLE"])
        S3(["CTD · SIDER · Reactome ..."])
    end

    subgraph ETL[" datasets/ "]
        direction TB
        SC["script.py"]
        GD["get_data.sh"]
        GD --> SC
    end

    subgraph Meta[" metadata/ "]
        direction TB
        MAP["mappings/\nGEN · CPD · DIS · CLL · TIS"]
        ONT["ontologies/\nDOID · GO · BTO · HPO"]
    end

    subgraph Processing[" code/kgraph/ "]
        direction TB
        UTIL["utils/\nmappers · ontology"]
        PROC["process_raw_data.py"]
        UTIL --> PROC
    end

    subgraph Embed[" code/embeddings/ "]
        direction TB
        EDGE["get_edges.py"]
        WALK["walks.py"]
        SKIP["mp2vec.py"]
        VAL["validation/"]
        EDGE --> WALK --> SKIP --> VAL
    end

    RAW[("graph/raw/\nedges by metaedge")]
    DONE[("graph/processed/\npropagated + depropagated")]
    EMB[("embeddings/\nnode vectors .h5")]

    Sources --> ETL
    MAP -.-> ETL
    ETL --> RAW --> Processing
    MAP -.-> Processing
    ONT -.-> Processing
    Processing --> DONE --> Embed --> EMB

    style Sources fill:#fef3e2,stroke:#bc6c25,stroke-width:1.5px,color:#6b4226
    style ETL fill:#f0f7e8,stroke:#588157,stroke-width:1.5px,color:#344e41
    style Meta fill:#e8f4f8,stroke:#2c7da0,stroke-width:1.5px,color:#184e77
    style Processing fill:#f0f7e8,stroke:#588157,stroke-width:1.5px,color:#344e41
    style Embed fill:#f0f7e8,stroke:#588157,stroke-width:1.5px,color:#344e41
    style RAW fill:#f3e8f9,stroke:#7b2d8e,stroke-width:1.5px,color:#4a1259
    style DONE fill:#f3e8f9,stroke:#7b2d8e,stroke-width:1.5px,color:#4a1259
    style EMB fill:#f3e8f9,stroke:#7b2d8e,stroke-width:2px,color:#4a1259

Loading

Quick start

cd datasets/gpsadb && python3 script.py   # process a dataset
python -m pytest -v                        # run tests

What's changed from upstream

• GPSAdb 2.0 dataset (7,665 gene perturbation experiments, 2,810 genes)
• Gene mappings regenerated from UniProt 2026_01 (+350 gene names)
• Provenance attributes and cell line links on perturbagen edges
• pytest suite for ETL and mapping generation

License

MIT (inherited from Bioteque, Copyright (c) 2022 SBNB)