ros2026-gcl-model

Computational model associated with:

Roš H., Xie Y., Sadeh S., Takarabe L.M., Silver R.A.
High-dimensional bidirectional population coding in pontine mossy fibre inputs to the cerebellar cortex
under review

Overview

This repository contains the computational model used to investigate how pontine mossy fibre population activity is transformed in the cerebellar granule cell layer.

The model is constrained by experimentally measured mossy fibre population activity and is designed to test how bidirectional, high-dimensional input patterns propagate through mossy fibre-to-granule cell connectivity. In particular, the simulations examine how different spatial and functional arrangements of mossy fibre inputs shape granule cell population responses.

The repository includes code for generating mossy fibre input spike trains from deconvolved activity, assigning those inputs to mossy fibres, constructing or loading connectivity, and running mossy fibre-granule cell simulations.

Scientific background

The ponto-cerebellar pathway provides the largest cortical input to the cerebellar cortex. In the associated study, population imaging revealed that pontine mossy fibre axons exhibit a high-dimensional bidirectional population code during spontaneous behaviour. The computational model in this repository was developed to test how these experimentally constrained input statistics interact with cerebellar input-layer architecture.

The simulations are used to explore how:

• mixed versus clustered organizations of mossy fibre inputs affect downstream granule cell activity
• opposing response types can be combined at the level of individual granule cells
• expansion from mossy fibres to granule cells shapes the dimensionality of the population code

Repository structure

• run_full_session_simulation.py — runs the full-session simulation pipeline by dividing a recording session into smaller temporal windows
• run_mixed_mf_grc_simulation.py — runs simulations with spatially mixed mossy fibre inputs
• run_clustered_mf_grc_simulation.py — runs simulations with spatially clustered mossy fibre inputs
• tools/ — utility scripts for input generation, behavioural classification, input assignment, and connectivity generation
• network_structures/ — mossy fibre-to-granule cell connectivity structures
• neuroml/ — NeuroML cell and synapse model files
• data/ — processed inputs required for the simulation workflow

Main workflow

A typical workflow consists of:

1. generating mossy fibre spike inputs from deconvolved activity
2. classifying and organizing inputs according to behavioural modulation
3. assigning spike trains to mossy fibres
4. simulating mossy fibre-to-granule cell network activity under mixed or clustered input configurations

The main entry point for running a session-level workflow is:

python run_full_session_simulation.py