Discovering Mechanistic Models of Neural Activity: System Identification in an in Silico Zebrafish
This work addresses the challenge of model verification in neuroscience, offering guidance for real-world neural recordings and a template for AI-driven scientific discovery, though it is incremental in applying existing methods to a new simulation context.
The researchers tackled the problem of verifying mechanistic models of neural circuits by creating an in silico testbed using larval zebrafish simulations as ground truth, and found that LLM-based tree search discovered predictive models that significantly outperformed established forecasting baselines, with structural priors enabling robust generalization and interpretable models.
Constructing mechanistic models of neural circuits is a fundamental goal of neuroscience, yet verifying such models is limited by the lack of ground truth. To rigorously test model discovery, we establish an in silico testbed using neuromechanical simulations of a larval zebrafish as a transparent ground truth. We find that LLM-based tree search autonomously discovers predictive models that significantly outperform established forecasting baselines. Conditioning on sensory drive is necessary but not sufficient for faithful system identification, as models exploit statistical shortcuts. Structural priors prove essential for enabling robust out-of-distribution generalization and recovery of interpretable mechanistic models. Our insights provide guidance for modeling real-world neural recordings and offer a broader template for AI-driven scientific discovery.