Knowledge-guided Contextual Gene Set Analysis Using Large Language Models.

Area of Science:

• Genomics
• Bioinformatics
• Artificial Intelligence in Medicine

Background:

• Gene set analysis (GSA) is crucial for interpreting disease genomic data by linking genes to biological processes.
• Conventional GSA methods often yield redundant or irrelevant pathways due to a lack of clinical context, complicating interpretation and reducing reproducibility.
• Manual interpretation of GSA results is time-consuming and subjective.

Purpose of the Study:

• To develop a novel AI-driven framework, cGSA, to enhance GSA by incorporating context-aware pathway prioritization.
• To improve the biological meaningfulness and clinical relevance of identified pathways from genomic data.
• To reduce the manual effort required for interpreting GSA results, thereby increasing reliability and reproducibility.

Main Methods:

• Developed cGSA, an AI framework integrating gene cluster detection, enrichment analysis, and large language models.
• Employed context-aware pathway prioritization to identify statistically significant and biologically relevant pathways.
• Benchmarked cGSA on 102 manually curated gene sets across 19 diseases and ten disease-related biological mechanisms.

Main Results:

• cGSA demonstrated over 30% improvement compared to baseline GSA methods in benchmarking studies.
• Expert validation confirmed cGSA's increased precision and interpretability of results.
• Case studies in melanoma and breast cancer highlighted cGSA's potential for uncovering context-specific insights.

Conclusions:

• cGSA offers a significant advancement over conventional GSA methods by integrating clinical context and AI.
• The framework enhances the identification of biologically meaningful pathways, supporting targeted hypothesis generation.
• cGSA improves the reliability, reproducibility, and interpretability of genomic data analysis in disease research.