Literature-scaled immunological gene set annotation using AI-powered immune cell knowledge graph (ICKG)

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Single-cell and spatial omics have identified novel gene sets in immunotherapy.
• Current methods for interpreting gene sets lack complexity, granularity, and accuracy.
• Existing literature contains rich immunological gene information but lacks semantic summarization for gene set analysis.

Purpose of the Study:

• To develop a novel approach for semantically summarizing immunological gene information.
• To create accurate and comprehensive annotations for gene sets discovered through omics studies.
• To overcome the limitations of traditional over-representation analysis for gene set interpretation.

Main Methods:

• Constructed immune cell knowledge graphs (ICKGs) by integrating over 24,000 published abstracts using large language models (LLMs).
• Validated ICKG quality using independent functional omics data (cytokine stimulation, CRISPR gene knock-out, protein-protein interactions).
• Developed an interactive website for ICKG-based gene set annotation and rationale visualization.

Main Results:

• ICKGs effectively integrate knowledge across peer-reviewed studies, enabling verifiable graph-based reasoning.
• Achieved rich, holistic, and accurate annotation of immunological gene sets.
• Successfully annotated previously unannotated gene sets and those used in clinical applications.

Conclusions:

• Immune cell knowledge graphs (ICKGs) offer a powerful solution for interpreting novel gene sets in immunology.
• LLM-driven knowledge graph construction enables accurate and comprehensive gene set analysis.
• The developed ICKG resource and interactive website facilitate deeper understanding of immunological gene functions and clinical applications.