GUI Network Builder: Modeling 20 synapses as 1

[Peaches205]

Is there anyway to model or compensate 20 sypnases with 20 connections into 1 connection with a single synpase. For instance, there are supposed to be 20 synapses on a neuron and an input fiber is supposed to stimulate/connect to each one individually. However, I am strictly working through the GUI network builder so importing 20 synapses for each type of synapse is a complete hassle. If there are 3 types of syanpses/recepotrs associated with an input fiber, then there would be 60 total syanpses to build. It is signficantly more efficient, time and computationally wise, for GUI network builder to make a single connection with a single synapse from the input fiber. I am open to any suggestions that could appropiately compensate!!

[ted]

The Network Builder is designed for prototyping simple networks that have only a few cells and a few synaptic connections (where "few" means maybe half a dozen at most; more than that would exceed the patience of most mortals). Also, the Network Builder doesn't allow more than a single instance of a particular synaptic class to be attached to the same "node" (the points at which NEURON computes membrane potential; for any section, that means the 0 and 1 ends and the centers of the section's segments).

Programming doesn't have those restrictions--it is easy to write code that creates as many cells as you like, and use as many NetCons as you like to drive any synaptic mechanism (the only limitation on the latter is that the synaptic mechanism must be capable of properly handling drive from multiple NetCons; NEURON's built-in synaptic classes ExpSyn1, ExpSyn2, and ExpSyn4 do that, and so do many user-developed synaptic mechanisms).

I'd suggest doing that with Python; here are some tutorials that can help you get started https://nrn.readthedocs.io/en/latest/tu ... index.html.

If the cell classes you created with the Network Builder have complex anatomical or biophysical properties, you might be able to save some time by exporting the cell classes from the Network Builder and then reusing them in your Python program.

revised by NTC on 6/30/2022

[Peaches205]

Thanks for getting back to me Ted, I really appreciate it. I wanted to make the program in a GUI for individuals who are not experienced in coding; however, it seems that is not possible with what I want. Is there a way I could download the network and cells in a hoc file and run it through an interpreter?

[ted]

First, I should mention that my previous reply was not as clear as it should have been. Please take a look at the revised version above.

I wanted to make the program in a GUI for individuals who are not experienced in coding

That is certainly doable. Just use hoc and/or Python to build the model network, then use NEURON's tools to make a user interface for running simulations and displaying results(and changing model parameters, if that would be useful). The user interface may include custom panels that contain highly customized controls, with labels that you choose, which perform very specific actions--in other words, exposing to the users convenient ways to take just those actions that you want them to be able to do. For specific examples of code that creates custom panels, see

https://nrn.readthedocs.io/en/latest/py ... dgets.html

. . . Is there a way I could download the network and cells in a hoc file and run it through an interpreter?

Not sure I understand what you're asing.

[Peaches205]

Ohhhhh, that is a smart idea. I did not even think of that, thank you. The only limitation is that the individuals using my program can only change what I allow them to (if they do not know how to code). Therefore, if there was something that needed to be changed on a fundamental level, I would have to do that myself, correct?

One last thing if you do not mind. If I wanted to make a passive node that would take a NetStim as an input, and connect the node to multiple synapses of other neurons, what would be the best technique. This node should not change the signal at all but rather just transmit it and reduce the number of connections needed (since eventually once I get one NetStim to work, I would connect 4 or 5 to that node). Therefore, instead of 4 or 5 NetStims connecting to 3 synapses each totaling up to 12-15 connections, a single node would only need a single connection from each netStim and 3 connections to the synapses of the neuron totaling up to 7 to 8.

[ted]

if there was something that needed to be changed on a fundamental level, I would have to do that myself

For any given model define A to be the set of "all things that can be done without writing more code." The problem is how to decide what belongs to A and what does not.

If I wanted to make a passive node that would take a NetStim as an input, and connect the node to multiple synapses of other neurons, what would be the best technique.

Given spike source S and targets T0, T1 . . . Tn, how would that be superior to simply connecting S directly to each of the targets?