Generating a Tolerogenic Cell Therapy Knowledge Graph from Literature.

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Tolerogenic cell therapies offer alternatives to immunosuppression for autoimmune diseases and transplant rejection.
• Keeping abreast of diverse research methodologies is crucial for advancing tolerogenic therapies.
• Knowledge graphs can link information sources via text mining for efficient data retrieval.

Purpose of the Study:

• To automatically generate a knowledge graph for tolerogenic cell therapy from biomedical literature.
• To develop a machine learning system (ICRel) for extracting cell-cytokine relations from abstracts.
• To facilitate the discovery of implicit or indirect relationships in published immunology studies.

Main Methods:

• Developed ICRel system using machine learning to extract cell-cytokine relations.
• Retrieved documents from PubMed and annotated abstracts for cell and cytokine entities.
• Generated cell-cytokine pairs co-occurring in sentences and identified meaningful relations.
• Constructed a knowledge graph where edges represent relations supported by documents.

Main Results:

• Created a knowledge graph with 647 cell-cytokine relations from 3,264 abstracts.
• The relation extraction module achieved an F-measure of 0.789.
• Manual evaluation of extracted relations demonstrated an accuracy of 0.615.

Conclusions:

• The ICRel system efficiently extracts implicit cell-cytokine relations from biomedical literature.
• The generated knowledge graph aids experts in identifying non-obvious connections in tolerogenic cell therapy research.
• This approach is more efficient than manual literature review for uncovering complex relationships.