Inferring brain-wide interactions using data-constrained recurrent neural network models.

CURBD infers directional inter-region currents across multiple brain regions from large-scale neural recordings, autonomously generating recurrent dynamics constrained by experimental data.

• Why it matters: Large brain-wide datasets require scalable models that can extract interpretable features of inter-region communication and support principled theories of coordination.
• What they did: The method uses data-constrained recurrent neural network models with current-based decomposition to infer functional interactions and directional currents, validated in simulated ground-truth networks.
• The result: CURBD accurately isolates inter-region currents in simulations and applies across larval zebrafish, mice, macaques, and humans, enabling untangling of brain-wide interactions linked to inter-area communication principles underlying behavior.

AI summary of the abstract · may contain errors · not medical advice · read the paper.