Questions about "stylized" model and build a new n

[huntadams]

Hello everybody,
I have two questions.
I started NEURON from "NEURON DEMO." I selected "stylized" model. I want to know which hoc file is corresponding to the stylized model, so I can read the file content and know the anatomical and biological properties of this model.

If I want to build a new neuron. Is a hoc file enough? I used the code in "The NEURON Simulation Environment Appendix", and saved it as a hoc file. However, it seemed not work when I loaded the file.

Thanks for the help!

[ted]

I started NEURON from "NEURON DEMO." I selected "stylized" model. I want to know which hoc file is corresponding to the stylized model, so I can read the file content and know the anatomical and biological properties of this model.

If you are interested in the demonstration program's models, reading the source code is
absolutely the wrong way to proceed. To properly create and destroy each individual
model plus associated graphical elements, it is loaded with administrative complexities
that are comprehension's deadly foes.

For the stylized model, you can use the old strategy of typing
topology()
at the xterm's oc> prompt to see how sections are assembled into the model's branched
architecture, and then type
forall psection()
to see hoc statements that, when executed, will set the anatomical and biophysical
properties of each section.

A better way to discover the biophysical specification is to use the Model View tool. To
start it, click on NEURON Main Menu / Tools / Model View
To learn about this tool, go to NEURON's Documentation page
http://www.neuron.yale.edu/neuron/docs
and click on the link
"ModelView: Compact display of parameters for NEURON models."

If I want to build a new neuron. Is a hoc file enough?

Depends on the contents of the hoc file.

I used the code in "The NEURON Simulation Environment Appendix", and saved it as a hoc file.

Do you mean the appendix of the expanded version of
Hines, M.L. and Carnevale, N.T.
The NEURON Simulation Environment
Neural Computation 9:1179-1209, 1997
?

However, it seemed not work when I loaded the file.

Exactly what did you do, and exactly what happened or didn't happen?

[mi2n]

Hello Ted,

Thanks for the directions, I understood the structure and mechanisms of "stylized" HH neuron model. I am curious, though, what were the references for building such a model? How did you decide the values for the membrane mechanisms and morphology? Were they just random values chosen from your experience? I couldn't find any pointers in the source files.

[ted]

How did you decide the values for the membrane mechanisms and morphology?

The same way that one would design a computational model of any other physical system. First determine the purpose of the model, which is usually to study some phenomenon of interest that occurs in a physical system. Then decide what aspects of the physical system should be represented in the model. This critical step involves abstraction and (a lot of) simplification from the infinite (and mostly unknown) complexity of all things physical. It requires knowledge of what is known or can reasonably be assumed about the physical system that is most pertinent to the phenomenon of interest, and it is guided by experience and judgement. The result is a conceptual model of the physical system that, in the opinion of the modeler, incorporates just enough detail to account for the phenomenon of interest. In other words, the conceptual model is a hypothesis about the phenomenon of interest.

The next step is to create the computational model, which is an implementation of the conceptual model expressed in computational form. Unlike the first step (formulation of the conceptual model), which involved simplification and abstraction, in going from conceptual model to computational model the modeler must not omit anything that was in the conceptual model, or add anything to the conceptual model. A close match between the conceptual model and the computational model is essential if the computational model is to be useful as a tool for gaining insight into the conceptual model.

[ted]

Continuing the discussion--

I did not implement this model, so regard everything that follows as informed speculation.

What was the purpose of this particular model? Presumably to use the phenomenon of spatial summation to illustrate certain features of NEURON. This required a cellular topology that had two or more dendrites, with a synapse attached to each.

Why include an axon? Well, what's the point of synaptic integration, if not to trigger the occasional spike that zips down an axon? So include an axon as a convenient and plausible "readout" of the result of spatial summation.

Why not use realistic cellular morphology and membrane biophysical properties, like those of my favorite class of neurons? That would violate the principle of parsimony (Occam's razor). In The notion of spatial summation is regarded as a general principle of cellular neurophysiology, so it should happen even in a very simplified model.

Why choose these particular anatomical and biophysical parameters? They're in the range of values observed in experiments on real cells. See the experimental literature, neuroelectro.org, recent textbooks on cellular neurophysiology.

[mi2n]

Thank you so much, for the detailed response on my query. I wanted to report a (approximate) technique to model ultrasound neuromodulation on NEURON models. As you recommended, I wanted to show it on a simplified model. I felt this "stylized" model is a good candidate, but I was wondering how these exact values ( For example, the equivalent diam of dendrite tree = 46 --> why not 50? ) are present in the model.

I got all my doubts clarified from your reply. And thanks for the "neuroelectro.org" -- that's a very useful page!