SCENE Framework Contextualizes Biomedical Knowledge into Scenario-Grounded Propositions

A new multi-agent AI framework called SCENE (Scenario Contextualization for Evidence-grounded Exploration) aims to bridge the gap between broad biomedical knowledge and specific experimental data. Developed by researchers at an undisclosed institution, SCENE uses a bi-level architecture: an upper level converts general knowledge into search directions grounded in dataset schemas, while a lower level executes multi-objective optimization to identify concrete propositions balancing evidential strength and data support. The framework treats knowledge contextualization as an iterative search process, enabling domain experts to inspect, replay, and validate propositions. This approach addresses the challenge where background knowledge is too general for direct mapping onto dataset variables, while data-driven patterns remain hard to interpret mechanistically. The paper is available on arXiv under ID 2605.27082.