Compiling molecular ultrastructure into neural dynamics

Authors: Konrad P. Kording, Anton Arkhipov, Davy Deng, Sean Escola, Seth G. N. Grant, Gal Haspel, Michał Januszewski, Narayanan Kasthuri, Nina Khera, Richie E. Kohman, Grace Lindsay, Jeantine Lunshof, Adam Marblestone, David A. Markowitz, Jordan Matelsky, Brett Mensh, Patrick Mineault, Andrew Payne, Joanne Peng, Xaq Pitkow, Philip Shiu, Gregor Schuhknecht, Sven Truckenbrodt, Joshua T. Vogelstein, Edward S. Boyden Date: 2026-03-26 Paper ID: arxiv:2603.25713

Summary

This paper proposes the concept of an Ultrastructure-to-Dynamics Compiler to bridge the gap between high-resolution molecular imaging of neuronal ultrastructure and understanding local neuronal physiology. The proposed compiler is a learned mapping designed to translate structural information, including synapse locations and molecular composition, into simulator-ready physiological parameters like synaptic efficacies and conductances. Successful implementation requires paired training data, combining joint ultrastructure imaging with physiological response measurements to perturbations. This capability promises to transform anatomical maps into predictive models of circuit dynamics, enabling better analysis of neural computation and forecasting the effects of biophysical interventions.

Key Contributions

• Introduction of the Ultrastructure-to-Dynamics Compiler, a learned mapping from molecularly annotated ultrastructure to physiological parameters.
• Framing the problem as needing paired training data of joint ultrastructure imaging and dynamical response experiments for model training.
• Enabling the shift from descriptive anatomical mapping to predictive modeling of local neuronal physiology and circuit dynamics.
• Providing a pathway to forecast the effects of interventions on neural circuits by using predicted physiological parameters.

Limitations

The primary limitation is the necessity for paired training data: simultaneously acquired ultrastructure from imaging and corresponding dynamical responses from physiological experiments, which is currently difficult to obtain.

Key Concepts

• Ultrastructure-to-Dynamics Compiler: A learned mapping that translates high-resolution, molecularly annotated neuronal ultrastructure data into simulator-ready, uncertainty-aware physiological parameters that govern neural dynamics.

Limitations

The primary limitation is the necessity for paired training data: simultaneously acquired ultrastructure from imaging and corresponding dynamical responses from physiological experiments, which is currently difficult to obtain.

Links

• ArXiv Abstract
• PDF