CNrun example

Supplementary material for the PLoS paper “Competition-based model of pheromone component ratio detection in the moth.” by Zavada A et al.

Note: the experimental setup described here uses CNrun 1.x, not the newer, Lua-powered version 2 (for that, check out this annotated script).

Experimental design

We used minimalistic, simplest conceivable neuronal network capable of detecting a certain target ratio of intensities of input components, by means of competition.

The model can be created:

In a tool such as neuroConstruct which generates a NeuroML network topology file (this file was actually used in our experiments; note however, that each of the clouds of ORNs are represented by a single Poisson oscillator). Use
```
 load_nml m.nml
```
to load the model.
Directly in cnrun with the following commands:
```
  new_model ratiocoding

  add_neuron Poisson ORNa
  add_neuron Poisson ORNb

  add_neuron HH LNa
  add_neuron HH LNb
  add_neuron HH LNz

  add_neuron HH PNi
  add_neuron HH PN

  add_synapse AB ORNa LNa .1
  add_synapse AB ORNa LNz .1
  add_synapse AB ORNb LNz .1
  add_synapse AB ORNb LNb .1

  # transverse
  add_synapse AB LNa LNb .1
  add_synapse AB LNb LNa .1
  # inbound
  add_synapse AB LNa LNz .1
  add_synapse AB LNb LNz .1
  # outbound
  add_synapse AB LNz LNa .1
  add_synapse AB LNz LNb .1
```
The model topology is now loaded in CNrun.
To actually run the simulation, you will need stimulation protocol files, and a CNrun script to properly set up all unit parameters. These files can be found in $prefix/share/aghermann/ratiocoding after you have done make install (or installed the .deb package); for your convenience, they are also available in this file. In the directory where you have the contents of setup tarball extracted, execute the batch shell script, like so:
```
$ ./batch 5 1.5
```
After 5 trials times 6 times a couple minutes per trial, you will have in each of the six newly created trial directories some .varx files, which record the E of so named neurons; these can be plotted with rational-plot-var. Similarly, the matrix data can be visulaised using rational-plot-sdf-static (like those shown in fig. 4) or, in a fancier manner, using rational-plot-sdf-interactive.

The batch script uses varfold (part of cnrun package) to ‘fold’ the spike density function values into a matrix, which is then used to calculate a cost function for a given trial.