ChAT Model

Introduction

This is an implementation of the spiking cerebellar granular layer model described in:

Taylor R Fore et al. "Acetylcholine modulates cerebellar granule cell spiking by regulating the balance of synaptic excitation and inhibition" Journal of Neuroscience 1 April 2020, DOI: https://doi.org/10.1523/JNEUROSCI.2148-19.2020

Code repository

The model is implemented using the Brian2 simulator for spiking neural networks, which is written in Python. The two scripts for running the simulation and generating the figures in the paper are:

• run_simulation.py - main script to run single instantiation of the model.
• figures.py - data analysis and visualization.

How to Use

Requirements

brian2==2.2.1
matplotlib
numpy>=1.19
pandas
python>=3.6.8,<3.8.0
scipy
scikit-learn

You can install these dependencies using either conda-lock.yml (recommended), environment.yml or requirements.txt file.

1. Install miniconda (recommended), conda, or mamba
   • We highly recommend changing conda's default solver to libmamba, see: https://www.anaconda.com/blog/conda-is-fast-now
2. Clone the repo and cd into the directory.
3. Install the dependencies into virtual environment, i.e. brian2

# clone the repo
$ git clone https://github.com/trfore/chatmodel.git
$ cd chatmodel

# Recommended, create 'brian2' environment
$ conda install conda-lock
$ conda-lock install --name brian2 lock_files/conda-lock.yml

# create env using environment.yml
$ conda env create -f environment.yml
# create env using requirements.txt
$ conda create --name brian2 --file requirements.txt --channel conda-forge

We recommend using the conda-lock.yml file, as it will install validated package versions tested against the following platforms.

Tested Platforms & Simulation Run Times

• Linux (Debian 10+, Ubuntu 20.04/22.04)
• macOS (10.13, 11, 12)
• Windows (10)

Simulation Run Times and Resource Utilization

• The full simulation takes ~ 42 minutes (2020 AMD 5600x, 32 GB) to 5 hours (2013 Intel I5-4258U, 8 GB) to run. This time is dependent on your system hardware.
• Unconstrained, the model uses a maximum of 27 GB of RAM and 2 physical cores, but will run on systems with less RAM as the simulation will write to cache and swap at the cost of a longer run time.
• The output files, *.npy, are collectively 328 MB in size. An archive is available on the release page (link: v1.1.0).

Running the scripts

Clone the repo and cd into the directory.

$ git clone https://github.com/trfore/chatmodel.git
$ cd chatmodel

All three drug conditions (control, GABAzine, muscarine) are built into the main script. To run the full simulation:

$ python run_simulation.py

The code saves all relevant variables as numpy arrays to the current directory for further analysis. Robust progress reporting is output to the console, but expect the entire simulation to take approximately 1 to 3 hours (depending on computer specs, see 'Tested Platforms & Simulation Run Times' for details).

To run analysis and generate the figures from the paper, type in terminal:

$ python figures.py

Figure windows will not appear, but pdf files of each figure will be saved in the working directory.

Environment Variables

You can change experiment parameters by setting environment variables, for example:

# Run the simulation for 50 trials
$ export CHAT_NUM_TRIAL=50
$ python run_simulation.py

Environment Variable	Default	Description	Required
CHAT_NUM_MF	315	int, number of mossy fiber inputs	No
CHAT_NUM_GRC	4096	int, number of granule cells	No
CHAT_NUM_GOC	27	int, number of golgi cells	No
CHAT_NUM_TRIAL	100	int, number of trials for the simulation	No
CHAT_RANDOM_WEIGHTS	True	boolean, use random weights for synaptic connections	No
CHAT_SEED_TRIAL	451	int, seed value for trial runs	No
CHAT_SEED_WEIGHT_MEANS	35	int, seed value for generating the random synaptic weight	No

Additional Notes

The paper was created using the following packages:

brian2==2.2.1
numpy==1.15.4
scipy==1.2.0
pandas==0.24.1
matplotlib==3.0.2
scikit-learn=0.19.2

The original requirements.txt file is available here: e69794d

---

Authors

• Nathan Taylor (https://github.com/taylorbn)
• Taylor Fore (https://github.com/trfore)

Maintainers

• Taylor Fore

---

References

• https://briansimulator.org/
• https://github.com/brian-team/brian2
• https://github.com/trfore/chatmodel

Anaconda & Mamba

• https://www.anaconda.com/download/
• https://docs.conda.io/en/latest/miniconda.html
• https://mamba.readthedocs.io/en/latest/installation.html
• https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html