TrajOnco Framework Uses Multi-Agent LLMs for Cancer Risk Prediction from Medical Records

TrajOnco Framework Uses Multi-Agent LLMs for Cancer Risk Prediction from Medical Records

Researchers have developed TrajOnco, a training-free multi-agent large language model framework designed for multi-cancer early detection using longitudinal electronic health records, according to a paper published on arxiv.org.

The framework uses a chain-of-agents architecture with long-term memory to perform temporal reasoning over sequential clinical events, generating patient-level summaries, evidence-linked rationales, and predicted risk scores. According to the research, TrajOnco was evaluated on de-identified Truveta EHR data across 15 cancer types using matched case-control cohorts to predict cancer diagnosis risk at 1 year.

In zero-shot evaluation, TrajOnco achieved AUROCs (Area Under the Receiver Operating Characteristic curve) of 0.64-0.80, performing comparably to supervised machine learning in a lung cancer benchmark while demonstrating better temporal reasoning than single-agent LLMs, according to the paper. The multi-agent design enabled effective temporal reasoning with smaller-capacity models such as GPT-4.1-mini.

The research states that TrajOnco’s output fidelity was validated through human evaluation. Additionally, the framework’s interpretable reasoning outputs can be aggregated to reveal population-level risk patterns that align with established clinical knowledge, according to arxiv.org.

The findings highlight the potential of multi-agent LLMs to execute interpretable temporal reasoning over longitudinal EHRs, advancing both scalable multi-cancer early detection and clinical insight generation, the paper concludes.