I have a model that I put together in XPP and I am in the process of inputting this model into NEURON. I am new to NEURON and I am rather confused about spatial settings (L and Diam). The model that I have does not have this data as it is a theoretical model. Is it possible to input models into NEURON without taking into account L and Diam?
I am rather new to this forum so please let me know if I posted this in the wrong section. Any help with this would be greatly appreciated.
Thanks,
J
Models Without Spatial Data

 Site Admin
 Posts: 5727
 Joined: Wed May 18, 2005 4:50 pm
 Location: Yale University School of Medicine
 Contact:
Re: Models Without Spatial Data
Need a little more information. Is this a completely abstract model, e.g. an artificial spiking cell with an "activation" state variable with a rest value of 0 and spike threshold of 1? A model of chemical reactions that occur in a well stirred volume? A model of channel gating? A single compartment model of a neuron that involves membrane capacitance and ionic currents? Or something else altogether? If it's a model that has already been published somewhere, the citation of the article would give me the information I need.
Re: Models Without Spatial Data
Thank you for the fast reply. The title of the paper that contains the model we are working with is: Distributed Neural Network Model for the Distinct roles of Medial and Lateral HVC in Zebra Finch Song Production. The link to the paper is as follows: https://www.math.fsu.edu/~bertram/paper ... JNP_17.pdf

 Site Admin
 Posts: 5727
 Joined: Wed May 18, 2005 4:50 pm
 Location: Yale University School of Medicine
 Contact:
Re: Models Without Spatial Data
Please answer yes or no: is the citation
Galvis D, Wu W, Hyson RL, Johnson F, Bertram R.
A distributed neural network model for the distinct roles of medial and lateral HVC in zebra finch song production.
J Neurophysiol. 118:677692, 2017
?
Given the complexity of the model described in that paper, your ability to replicate the model will depend critically on the availability of the original source code (because the likelihood of a typographical error or simple omission of one or more critical details will be high). I see that the senior author (Bertram) has made some code available on his own web pages at FSU e.g. https://www.math.fsu.edu/~bertram/softw ... ng/JNP_17/ so maybe that will be sufficient to get you started.
After you answer the above question, we can get into specifics.
Galvis D, Wu W, Hyson RL, Johnson F, Bertram R.
A distributed neural network model for the distinct roles of medial and lateral HVC in zebra finch song production.
J Neurophysiol. 118:677692, 2017
?
Given the complexity of the model described in that paper, your ability to replicate the model will depend critically on the availability of the original source code (because the likelihood of a typographical error or simple omission of one or more critical details will be high). I see that the senior author (Bertram) has made some code available on his own web pages at FSU e.g. https://www.math.fsu.edu/~bertram/softw ... ng/JNP_17/ so maybe that will be sufficient to get you started.
After you answer the above question, we can get into specifics.
Re: Models Without Spatial Data
Yes, that is the correct citation. Since I collaborate with Bertram, I have access to the XPP source code. NEURON has additional features that XPP does not and NEURON seems to be the gold standard so we plan on transferring all of our models into NEURON.
I’ve given an example of one of our partial models: See code below:
https://www.math.fsu.edu/~bertram/cours ... /HVCRA.ode
I’ve given an example of one of our partial models: See code below:
https://www.math.fsu.edu/~bertram/cours ... /HVCRA.ode
Re: Models Without Spatial Data
I misread your response. My apologies. Yes the code you found on Bertram’s page is a good starting example and I managed to successfully import it into NEURON but the model is affected by Diam and L. My familiarity with programming in NEURON is rather limited so I’m not familiar with importing purely theoretical models without Diam and L which would enable us to transition our source code from XPP to NEURON. We essentially don’t have any geometry information on our HVC cells that we are modeling.

 Site Admin
 Posts: 5727
 Joined: Wed May 18, 2005 4:50 pm
 Location: Yale University School of Medicine
 Contact:
Re: Models Without Spatial Data
I just now emailed you int.zip, which contains a NEURON implementation of INT.ode, the interneuron model described by Galvis et al.. That email also includes some notes about how to proceed in porting from XPP to NEURON. For the sake of others who read this thread, the process involves dealing with these tasks:
0. Deciding where to start.
1. Transforming model parameters from absolute units to density units.
2. Restructuring code.
A. Making it modular to facilitate debugging and revision. This means separating model specification code (stuff that specifies the aspects of the biological system that are represented in the model) from code that specifies experimental conditions, instrumentation (signal sources and recording of results), user interface (display of results), and simulation control (integration method, time step, duration of simulation etc.).
B. Moving cell models from top level to instances of cell classes.
3. Specifying ion channel properties via NMODL.
4. Taking advantage of NEURON's standard run system for initialization and simulation flow control
5. Taking advantage of NEURON's GUI for convenient display of simulation results.
6. Debugging.
7. Verifying that the ported model replicates the results produced by the original XPP implementation.
With regard to specifying section L and diam, the XPP code provides a key clue: total membrane capacitance is 30 pF. Membrane capacitance is generally assumed to be 1 uF/cm2, which is equivalent to 1 pF/100 um2. Total membrane capacitance of 30 pF means surface area should be 30*100 = 3000 um2. Since this is a single compartment model, you could make L and diam be anything that satisfies PI*L*diam = 3000. However, for purposes of illustration, it's nice to have L and diam of a single compartment model be roughly comparable. Also, it's best for L to have a "nice" value. Since L*diam ~ 1000, might as well assume L to be 30 um, and then diam = 3000/(PI*L).
As far as ion channel densities are concerned, it helps to know that a single compartment with surface area 100 um2 and channel density of 1 S/cm2 has a net conductance of 1 nS.
0. Deciding where to start.
1. Transforming model parameters from absolute units to density units.
2. Restructuring code.
A. Making it modular to facilitate debugging and revision. This means separating model specification code (stuff that specifies the aspects of the biological system that are represented in the model) from code that specifies experimental conditions, instrumentation (signal sources and recording of results), user interface (display of results), and simulation control (integration method, time step, duration of simulation etc.).
B. Moving cell models from top level to instances of cell classes.
3. Specifying ion channel properties via NMODL.
4. Taking advantage of NEURON's standard run system for initialization and simulation flow control
5. Taking advantage of NEURON's GUI for convenient display of simulation results.
6. Debugging.
7. Verifying that the ported model replicates the results produced by the original XPP implementation.
With regard to specifying section L and diam, the XPP code provides a key clue: total membrane capacitance is 30 pF. Membrane capacitance is generally assumed to be 1 uF/cm2, which is equivalent to 1 pF/100 um2. Total membrane capacitance of 30 pF means surface area should be 30*100 = 3000 um2. Since this is a single compartment model, you could make L and diam be anything that satisfies PI*L*diam = 3000. However, for purposes of illustration, it's nice to have L and diam of a single compartment model be roughly comparable. Also, it's best for L to have a "nice" value. Since L*diam ~ 1000, might as well assume L to be 30 um, and then diam = 3000/(PI*L).
As far as ion channel densities are concerned, it helps to know that a single compartment with surface area 100 um2 and channel density of 1 S/cm2 has a net conductance of 1 nS.