Automated Standardization of Legacy Biomedical Metadata Using an Ontology-Constrained LLM Agent

The article presents a novel LLM-based system for standardizing legacy biomedical metadata by integrating real-time queries to standard reporting guidelines and authoritative terminology services. The evaluation on 839 records from the Human BioMolecular Atlas Program demonstrates that this approach significantly enhances prediction accuracy compared to traditional LLMs, highlighting its potential for improving dataset findability and interoperability in biomedical research. This method addresses the challenge of producing FAIR datasets by enabling dynamic access to machine-actionable metadata standards, which is crucial for practitioners working with diverse biomedical datasets.