The paper introduces MEDA, an LLM- and symbolic-regression-powered agentic framework for discovering ordinary differential equation models of biological and biologically inspired dynamical systems. MEDA retrieves background knowledge, defines admissible variables, generates mechanistic constraints, proposes candidate equations, and fits and evaluates them. The authors evaluate canonical model retrieval, reasoning-based extrapolation to unseen variants, and open-ended discovery, with and without experimental data. They report strong structural recovery in retrieval and extrapolation settings and biologically plausible discovery-oriented models. Ablation and robustness analyses identify knowledge-guided formalization and mechanistic constraints as load-bearing components, while numerical fitting alone can produce biologically incorrect but trajectory-compatible equations.
No heat snapshots are available in the last 24 hours.