Network bursts in cultured NN result from different adaptive mechanisms (Masquelier & Deco 2013)

 Download zip file 
Help downloading and running models
It is now well established that cultured neuron networks are spontaneously active, and tend to synchronize. Synchronous events typically involve the whole network, and have thus been termed “network spikes” (NS). Using experimental recordings and numerical simulations, we show here that the inter-NS interval statistics are complex, and allow inferring the neural mechanisms at work, in particular the adaptive ones, and estimating a number of parameters to which we cannot access experimentally.
1 . Masquelier T, Deco G (2013) Network bursting dynamics in excitatory cortical neuron cultures results from the combination of different adaptive mechanisms PLoS ONE 8(10):e75824 [PubMed]
Model Information (Click on a link to find other models with that property)
Model Type: Realistic Network;
Brain Region(s)/Organism:
Cell Type(s): Cochlear nucleus pyramidal (fusiform) cell; Abstract integrate-and-fire adaptive exponential (AdEx) neuron;
Channel(s): I_AHP;
Gap Junctions:
Receptor(s): AMPA; NMDA; Glutamate;
Transmitter(s): Glutamate;
Simulation Environment: Brian; Python;
Model Concept(s): Bursting; Temporal Pattern Generation; Synchronization; Short-term Synaptic Plasticity; Facilitation; Depression; Spike Frequency Adaptation;
Implementer(s): Masquelier, Tim [timothee.masquelier at];
Search NeuronDB for information about:  Cochlear nucleus pyramidal (fusiform) cell; AMPA; NMDA; Glutamate; I_AHP; Glutamate;
This readme and the code were contributed by Timothee Masquelier
Aug 2013

This code was used in:

Masquelier T and Deco G (2013) Network bursting dynamics in excitatory
cortical neuron cultures results from the combination of different
adaptive mechanisms. PLoS ONE

Feel free to use/modify but please cite us if appropriate.

We use the Brian simulator described in:
Goodman D, Brette R (2008) Brian: a simulator for spiking neural
networks in python. Front Neuroinformatics 2:5
and available at:

This code has been tested with:
   - Brian 1.4.0
   - Python 2.6
   - Mac OS and Linux

Main file: (should be called like that "python -i")
Calls to set the parameters (see comments there), launches
the simulation and plots the results.

The current values in corresponds to the baseline simulation
in the paper (1 min of simulated time).

Many more options are provided (with no guaranty), eg sparse
connectivity, inhomogeneous weights, STDP, adaptive thresholds,
inhibitory neurons etc.

Note: the AMPA and NMDA inputs are computed manually at each time step
using a custom @network_operation (NOT using Brian built in
Connection).  For the simplified case of full connectivity with
homogeneous, non-plastic, weights, these contributions are the same
for all neurons, which enormously simplifies computation.

Python files:
--------------------                        main script                       contains all the parameters        Brian file to handle both a refractory
                               period and a user-defined reset
                               function            handles Short Term Plasticity                  graphical output

Output files: (in ./data/)

spike.%random seed%.%dump number%.mat Output spikes. Format nx2
                                      matrix, first column is neuron
                                      indexes, second is spike times.

dump.%random seed%.mat                Contains g_a ("adapt"), membrane
                                      potentials ("pot_e"),firing
                                      rates ("rate"), facilitation
                                      ("u") and depression ("x")

Loading data, please wait...