NEST models and simulation paradigms

States

We consider here two types of mouse experiments that the canonical circuit and each derived model is reproducing in simulation:

  • The in-vitro experiments of a slice of isolated mouse cerebellar tissue.

  • The awake experiments of behaving mouse.

Additionally, the awake state is derived from the in-vitro state, as only some cells and connections receive modifications.

Neuron models

All the default NEST neuron models are available for simulation with BSB. Additionally, BSB allows for the different populations’ parameters to be described as distribution or plain values:

simulations:
    simulation_name:
        simulator: nest
        seed: 1234
        cell_models:
          granule_cell:
            model: eglif_cond_alpha_multisyn
            constants:
              t_ref: 1.5
              I_e: -0.888
              Vinit:
                distribution: normal
                mean: -62.0
                std: 20.0

The distribution parameter correspond to one of the functions of the NEST random module. It can be useful to randomize certain parameters such as the initial membrane potential to avoid synchrony of activity at the start of the simulation.

Note

As in the previous configuration snippet, you can set the NEST random seed with BSB using the seed parameter.

Parrot neurons

Parrot neurons are the most basic neuron model of NEST. They basically emits one spike for every incoming spike they receive. We are using them here to represent the mossy fibers and glomerulus population as these fibers only transmit spikes coming from other regions of the brain.

Geminiani EGLIF

This model is based on the work of Geminiani et al. [1] [3] [4]. Details about the model are available in this page.

De Grazia EGLIF

This model is based on the work of De Grazia et al. [2]. Details about the model are available in this page.

Synapse models

Static synapse

By default, NEST static synapses are used to connect the different neurons together. This model only transmit spikes as weights to postsynaptic neurons after a provided delay.

Tsodyks Markram Synapse

The Tsodyks-Markram synapse model implements synaptic short-term depression and short-term facilitation according to Tsodyks et al. [5] and Fuhrman et al. [6]. This connection model merely scales the synaptic weight, based on the spike history parameters of the kinetic model.

Two versions of this model are available on NEST:

Synapse parameters

Synaptic parameters depend on the neuron model chosen for the simulation. Find more information on these parameters in the neuron model associated pages.

Simulation paradigms

Different configuration files are available to reproduce experiments with the cerebellar cortex circuit. As for the circuit reconstructions, the simulations are based on a basic paradigm (see the following section) and can be extended with simulation extensions.

To analyze the spiking results of the following simulations, for each neuron population, we define:

  • the mean firing rate as the mean of each of its (spiking at least once) neurons’ total number of spikes over the simulation time, expressed in Hz.

  • the mean Inter-Spike intervals (ISI) as the mean of each of its (spiking at least twice) neurons mean duration between each of its pair of consecutive spike, expressed in ms.

Basal activity

The basal activity configuration file basal.yaml implements to the default activity of the cerebellar cortex circuit.

This simulation is set to last 5000 ms (with a 0.1 ms timestep) during which the neurons are only stimulated with background noise represented as a 4 Hz Poisson spike generator on the mossy fibers population. Each neuron spiking activity is additionally recorded.

This simulation should demonstrate the activity of the network in a stable state. The results of this simulation serves as a baseline for the following ones.

Mossy fiber stimulus

The mossy fiber stimulus configuration file stimulus_mossy.yaml is derived from the basal_activity but adds a stimulus of the mossy fibers (see Basal activity).

On top of the basal paradigm, we introduce here a stimulus represented as a 150 Hz Poisson spike generator between 1200 and 1250 ms. This latter targets the mossy_fibers that are within a sphere of radius 90 \(\mu m\) and a center at (150.0, 65.0, 100.0). This mimic the integration of an input on the mossy fibers.

Simulation results

Simulation results depend on which cells types are included in the canonical circuit, the state, the simulation paradigm, and the neuron and synapse models used. The results for the eglif_cond_alpha_multisyn with static_synapse are available in this page.

References