GeneKnow: An auditable AI framework for source-grounded biological evidence synthesis

Area of Science:

• Biomedical Research
• Bioinformatics
• Artificial Intelligence in Science

Background:

• Interpreting gene function in specific biological contexts is crucial for advancing biomedical research.
• Manual literature review for gene function is time-consuming and labor-intensive.
• Existing AI tools often lack reliability and can produce fabricated citations or hallucinations.

Purpose of the Study:

• To develop a source-grounded framework, GeneKnow, for reliable and traceable literature synthesis.
• To leverage generative AI within a controlled hybrid workflow for accurate gene function interpretation.
• To address the limitations of manual review and current AI tools in biomedical literature synthesis.

Main Methods:

• Developed GeneKnow, a framework integrating generative AI models.
• Implemented a controlled hybrid workflow for literature synthesis.
• Conducted systematic benchmarking against mainstream AI web-interface tools.

Main Results:

• GeneKnow produces reliable, traceable literature syntheses supported by authentic citations.
• The framework effectively minimizes hallucinations in gene function interpretation.
• GeneKnow demonstrated superior performance compared to existing AI tools in generating context-specific gene function syntheses.

Conclusions:

• GeneKnow offers a trustworthy solution for synthesizing gene function information from scientific literature.
• The framework enhances the efficiency and reliability of biomedical research by reducing manual effort and AI errors.
• GeneKnow represents a significant advancement in applying AI for accurate scientific literature analysis.