Knowledge Graph Framework to Generate Hypotheses for Natural Product-Drug Interactions

NP-KG is a graph framework that creates a biomedical knowledge graph (KG) to identify and generate mechanistic hypotheses for pharmacokinetic natural product-drug interactions (NPDIs). NP-KG uses the PheKnowLator ecosystem to create an ontology-grounded KG. It then uses two relation extraction systems to extract triples from full texts of natural product-related scientific literature to create a literature-based graph, and integrates the nodes and edges in the ontology-grounded KG.

NP-KG: Merged PheKnowLator KG and literature-based graph with 30 natural products.

Ontology-grounded KG: PheKnowLator KG with a few additional data sources.

Literature-based Graph: Literature-based graph constructed from scientific literature with relation extraction systems (SemRep and INDRA/REACH) and closure operations.

1. Clone the repository or download all files.
2. Install all required packages. Requires Python>=3.6.

python -m pip install -r requirements.txt

1. Download the knowledge graph and node labels files from Zenodo and add to local folder - resources/knowledge_graphs. NP-KG is available as TSV file with triples and NetworkX multidigraph (gpickle files).

• Merged KG: includes merged PheKnowLator KG and literature-based graph. Download this file if you do not know which KG to use.
  • Filename: NP-KG_v3.0.0.tsv
  • Filename: NP-KG_v3.0.0.gpickle
• PheKnowLator KG: includes full instance-based build of the PheKnowLator KG. See PheKnowLator for more details.
  • Filename: PheKnowLator_v3.1.2_full_instance_inverseRelations_OWLNETS_NetworkxMultiDiGraph.gpickle

• Download nodeLabels_v3.0.0.tsv file with all node labels for the merged KG.
• Download nodeTypes_v3.0.0.tsv file with node types for all nodes in the merged KG.

2. See evaluation-scripts for examples of queries and path searches.

Note: The download link also contains the KGs as gpickle and ntriples files with the same nodes and edges that can be loaded for other applications.

The Graph Representation Learning library GRAPE provides efficient graph embeddings. To load NP-KG (version 3.0.0) in GRAPE, use the from_csv function and TSV files mentioned above:

npkg = Graph.from_csv(
            node_path=<TSV node types filename>,
            node_list_node_types_column_number=1,
            nodes_column_number=0,
            node_list_separator='\t',
            node_list_header=True,
            edge_path=<TSV version of KG filename>,
            edge_list_separator='\t',
            edge_list_header=True,
            edge_list_edge_types_column_number=1,
            sources_column_number=0,
            destinations_column_number=2,
            weights_column_number=3,
            directed=True,
            verbose=True
        )

NP-KG (v1.0.1) can also be loaded with as below. See NP-KG Grape Animation tutorial for details.

pip install grape -U
from grape.datasets.zenodo import NPKG
graph = NPKG(directed=True)
graph

See wiki for details of data sources, construction, use cases, and evaluation.

Get in touch through GitHub issues, discussion, or email!

NP-KG Publication

Taneja SB, Callahan TJ, Paine MF, Kane-Gill SL, Kilicoglu H, Joachimiak MP, Boyce RD. Developing a Knowledge Graph Framework for Pharmacokinetic Natural Product-Drug Interactions. Journal of Biomedical Informatics. 2023. DOI: doi.org/10.1016/j.jbi.2023.104341.

AMIA Informatics Summit poster

Taneja SB, Ndungu PW, Paine MF, Kane-Gill SL, Boyce RD. Relation Extraction from Biomedical Literature on Pharmacokinetic Natural Product-Drug Interactions. Poster presentation, AMIA Informatics Summit 2022; March 21-24, 2022.

ISMB Conference Abstract and Related Files

Taneja SB, Callahan TJ, Brochhausen M, Paine MF, Kane-Gill SL, Boyce RD. Designing potential extensions from G-SRS to ChEBI to identify natural product-drug interactions. Intelligent Systems for Molecular Biology/European Conference on Computational Biology (ISMB/ECCB), 2021. https://doi.org/10.5281/zenodo.5736386

Publication

@article{taneja_developing_2023,
	title = {Developing a {Knowledge} {Graph} for {Pharmacokinetic} {Natural} {Product}-{Drug} {Interactions}},
	volume = {140},
	issn = {1532-0464},
	url = {https://www.sciencedirect.com/science/article/pii/S153204642300062X},
	doi = {10.1016/j.jbi.2023.104341},
	language = {en},
	urldate = {2023-03-23},
	journal = {Journal of Biomedical Informatics},
	author = {Taneja, Sanya B. and Callahan, Tiffany J. and Paine, Mary F. and Kane-Gill, Sandra L. and Kilicoglu, Halil and Joachimiak, Marcin P. and Boyce, Richard D.},
	year = {2023},
}

Zenodo Dataset

@dataset{taneja_sanya_bathla_2024_12536780,
  author       = {Taneja, Sanya Bathla},
  title        = {{NP-KG: Knowledge Graph for Natural Product-Drug 
                   Interactions}},
  month        = jun,
  year         = 2024,
  publisher    = {Zenodo},
  version      = {3.0.0},
  doi          = {10.5281/zenodo.12536780},
  url          = {https://doi.org/10.5281/zenodo.12536780}
}

This work is supported by the National Institutes of Health National Center for Complementary and Integrative Health Grant U54 AT008909.