SL-MERK: Synthetic Lethality Mechanism Explainer based on GraphRAG and Knowledge Graph.

Area of Science:

• Biomedical Informatics
• Computational Biology
• Artificial Intelligence in Oncology

Background:

• Synthetic lethality (SL) is a promising cancer therapy strategy targeting cancer-specific vulnerabilities.
• Limited understanding of SL mechanisms hinders clinical translation despite advances in screening technologies.
• Artificial intelligence, particularly large language models (LLMs), offers new potential for elucidating complex biological mechanisms.

Purpose of the Study:

• To develop a novel computational framework for explaining synthetic lethality (SL) mechanisms.
• To integrate Graph Retrieval-Augmented Generation (GraphRAG) with knowledge graphs for enhanced SL mechanism explanation.
• To leverage LLMs for generating comprehensive and interpretable natural language explanations of SL.

Main Methods:

• Proposed SL-MERK framework combining GraphRAG for extracting SL interaction patterns from literature.
• Integrated knowledge graphs to provide rich mechanistic context.
• Utilized LLMs for generalization and natural language generation of explanations.

Main Results:

• SL-MERK framework successfully generated interpretable natural language explanations of SL mechanisms.
• Experimental evaluations showed SL-MERK significantly outperformed GPT-4 and other LLMs in explanatory performance.
• The framework effectively combined literature-derived interaction patterns with knowledge graph information.

Conclusions:

• The proposed SL-MERK framework provides a powerful AI-driven approach for understanding synthetic lethality mechanisms.
• This method enhances the interpretability and comprehensiveness of SL mechanism explanations.
• SL-MERK represents a significant advancement in computational approaches for cancer therapy development.