The abstract is read at the publisher; the summary is JClub's.
An agentic ReAct LLM system for TCGA clinical data achieves 93.4% overall accuracy across 440 benchmark queries, enabling auditable natural-language statistical analysis.
- Why it matters: TCGA’s complex, heterogeneous clinical files make access and programming difficult, limiting accurate, reproducible querying and analysis for clinicians and researchers.
- What they did: The system uses an LLM ReAct agent selecting from eight tools (e.g., extraction, descriptive stats, Kaplan-Meier with log-rank tests, hypothesis testing) and verifies outputs against the curated TCGA Pan-Cancer Clinical Data Resource (CDR).
- The result: It reaches 93.4% accuracy versus 87.1% for a rule-based pipeline, 81.8% for a single-pass LLM, and 66.9% for retrieval-augmented generation; for 26 CDR-missing fields, it answers 100% versus 3.8% with CDR-only.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.