neurostar / matlab-utils Goto Github PK

Created after many years of graduate school in computational neuroscience, and covering a wide range of tasks from data analysis, to figure preparation with a large amount of system-connecting in between.

I am mostly adding putting this up for the sake of the Mathoc subfolder, which contains a relatively complete and functional bridge between matlab and the NEURON simulation environment. Rather than explain lugubriously, an example may instead be more illustrative of how it works and what it does:

init_hoc;
tstop = 5;
hoc_variable(tstop);
dt = .0125;
hoc_variable(dt);


call('load_file', "utilities.hoc");
call('load_file', "savers.hoc");
hoc('create soma');
neurons = {};
nNeurons = 50;
inhPerc = .2;

inhibitoryFlag = (1:nNeurons)<(nNeurons.*inhPerc);

for i=1:nNeurons
  neurons{i} = instantiate('IntFire4',.5)
end

connections = {};
for i=1:length(neurons)
  for j=1:length(neurons)
	if i~=j
	  if inhibitoryFlag(i)
		w = -.03;
	  else
		w = .01;
	  end
	  connections{i,j} = instantiate('NetCon',neurons{i},neurons{j},0,0,w);
	end
  end
end

for i=1:nNeurons
  stim = instantiate('NetStim');
  hoc_set(stim,'interval',1);
  hoc_set(stim,'number',1000);
  hoc_set(stim,'noise',1);

  netcon = instantiate('NetCon',stim,neurons{i},1,0,5);

end

for i=1:nNeurons
  recording1 = instantiate('VoltageTraceSaver',...
						  'a',...
						  -1,...
						  -1,...
						  code('utils.nil'),...
						  sprintf('/tmp/recording%d.data',i),neurons{i},'m');
end

simulation_loop(1);

for i=1:nNeurons
  delete(sprintf('/tmp/recording%d.data',i));
end

do_hoc(1);
data = [];
for i=1:nNeurons
  data = [data; load(sprintf('/tmp/recording%d.data',i))];
end
t = linspace(0,tstop,length(data));
close all
plot(t,data);

The above piece of code uses NEURON, via Matlab or Octave, to create a dead stupid model of a micro-column-like neural network. Fifty neurons are created, and around 20% are assigned inhibitory status. Then synaptic connections are created between all neurons depending on that status. Finally, excitatory input is created via a series of NetStim objects connected to each neuron, and objects for recording the voltages of the neurons are created. do_hoc initiates the simulation, and then the data from it is loaded and plotted.

Documentation is sparse right now, but the above example shows you how to do most everything you'd want to do. The system needs to know where your Neuron executable (nrniv) is located. You tell it by specifying the global variable hoc_exe_name__.

This whole system is based on Matlab compiling a hoc file describing the simulation, and then running it when do_hoc is called. You can check the file by looking at the global variable hoc__.

There is a lot of stuff going on behind the scenes with file names and so forth, and a lot of utilities on the hoc side to make things easier. I'll talk about them as I document them.

Finally, you can't use this package without having my whole matlab repository on the path somewhere (including sub-directories).