Astrocytes learn to detect and signal deviations from critical brain dynamics

This package is a Tensorflow implementation of the astrocyte calcium response to changes in neural activity dynamics around a critical phase transition.

The paper has been accepted at Neural Computation, available here.

Citation

Vladimir A. Ivanov and Konstantinos P. Michmizos. "Astrocytes learn to detect and signal deviations from critical brain dynamics." Neural Computation (2022).

@article{ivanov_2022,
author = {Ivanov, Vladimir A. and Michmizos, Konstantinos P.},
title = {Astrocytes learn to detect and signal deviations from critical brain dynamics},
year = {2022},
journal = {Neural Computation}
}

Software Installation

• Python 3.6.9
• Tensorflow 2.1 (with CUDA 11.2 using tensorflow.compat.v1)
• Numpy

Usage

This code performs the following functions:

1. Generate the attractor network (Ising model) with clustered couplings
2. Evaluate astrocyte Calcium response to change in neural dynamics
3. Visualize simulation output

1. Generate attractor network

To generate the attractor network (Ising model) with clustered couplings, enter the following command:

   python ICA_ising_model_builder.py

with inputs:

• --model_num : Integer assigned to model created.

This will create a file formatted as ISING_Model_X in subfolder '/Ising_models/Model_X/, where X is the model number set by input --model_num.

2. Evaluate astrocyte calcium response

To run a simulation with biophysical astrocyte driven by neural activity with dynamics transitioning from T1 to T2, enter the following command:

  python ICA.py

with inputs:

• --ver_num : Integer assigned to simulation run.
• --ising_model_num : Integer assigned to Iisng model used.
• --t1 : Float represeting start dynamics of simulation. Astrocyte weights are initialized on neural activity from these dynamics.
• --t2 : Float represeting ending dynamics of simulation. This controls the dynamics to which the Ising model transitions to with astrocyte already initialized on t1.

This will create a file formatted as ICA_Data_ver_X in subfolder '/dataFiles/ver_X/, where X is the simulation version number set by input --ver_num.

3. Visualize output

To visualize the output from the simulation file in the last step, enter the following command:

  python ICA_Plotter.py

with inputs:

• --ver_num : Simulation verion number from which to plot.
• --ising_model_num : Integer assigned to Iisng model used.

This will create the following image files in subfolder /dataFiles/ver_X where X is the simulation number.

• Astrocyte Frequency Response.png : shows time series of astrocyte calcium concentration response to Ising synaptic activity transition from t1 to t2.
• Synaptic Rate Distributions.png : shows distribution of synaptic rates for t1 and t2.
• Synaptic Rates.png : shows a 2D visualization of synaptic rates.

Other files

ICA_ising.py : Contains class for creating and running Ising model.

ICA_astrocyte.py : Contains class for initializing and running astrocyte model.

ICA_coupling_pattern.py : Contains class for generating surface over spin coordinates used for creating Ising model.

ICA_support_lib.py : Contains set of house keeping support functions.

Example commands

    python ICA_ising_model_builder.py --model_num=1

    python ICA.py --ver_num=10 --ising_model_num=1 --t1=2.2 --t2=3.2

    python ICA_Plotter.py --ver_num=10 --ising_model_num=1   

Create an Ising model with clustered couplings using run file: * 'ICA_ising_model_builder.py'

2. Launch ICA simulation using the created Ising model using run file:

   • 'ICA.py'

3. Generate visualize outputs from simulation using run file:

   • 'ICA_Plotter.py'