Model of the cerebellar granular network

This is a physiologically detailed model of the cerebellar granular network, used for

Sudhakar, S.K., Hong, S., Raikov, I., Publio, R., Lang, C., Close, T., Guo, D., Negrello, M., De Schutter, E., 2017. Spatiotemporal network coding of physiological mossy fiber inputs by the cerebellar granular layer. PLOS Comput Biol 13, e1005754. doi:10.1371/journal.pcbi.1005754

This model runs on the NEURON simulation platform with an MPI-based parallel running feature enabled.

How to run the model

Steps to run a simulation using our model are:

  1. Install BREP,
  2. Prepare a parameter file and mossy fiber spike times,
  3. Construct and run a simulation.

Below are details about each step.

BREP installation

BREP is a program to compute the connectivity between parallel fibers and Golgi cells. Compiling this requires the latest version of chicken-scheme, working f2c libraries, and MPI environment.

After installing these, go to brep directory, modify variables in brep-deploy-linux.sh, and run it to install BREP.

Preparation of a parameter file and mossy fiber spike times

The simulation will look for a directory that contains a parameter files and also files about mossy fibers. We prepared an example in params/set3005, which will generate a result in Figure 2 of the paper ("On patch network oscillation"). In this directory, you will find:

  1. Parameters.hoc, which has various parameters including run time, time step, density of neurons, etc.,

  2. MFCr.dat, which has coordinates of mossy fibers,

  3. GLpoints.dat, which has coordinates of mossy fiber rosettes,

  4. activeMfibres1.dat, which has a list of active (firing) mossy fibers,

  5. l.dat, which lists the number of spikes for each active mossy fiber,

  6. datasp.dat, which contains spike times of active mossy fibers in milliseconds.

For generation of *.dat files, please see example MATLAB scripts in mossy_fiber_utils.

Construction and running of a simulation

A simulation is constructed and run in four steps.

  1. All of mechanism (*.mod) files are compiled,
  2. Cell populations are generarted,
  3. Connectivity between Golgi cells and parallel fibers is computed,
  4. The simulation is launched.

The first two and last steps are handled by main.py, and the third step requires installed BREP. For more details, take a look at an example batch script to run all the steps, batch_slurm_start.sh, which first compiles mod files and then calls population_init.slurm, pf_goc_projection.slurm, and simulation_job.slurm in order for step 2, 3, and 4, respectively. These scripts are written for cluster computers managed by SLURM system, and would need to be changed accordingly in different systems.


Written by Sungho Hong, Computational Neurosience Unit, Okinawa Institute of Science and Technology, Japan

September 16, 2017

Correspondence: , Computational Neurosience Unit, Okinawa Institute of Science and Technology, Japan