Cell Builder: simulation changes after saving and reloading

[horstmann]

Hello!

I' working with the cell builder on an anatomically detailed model. At the moment I'm trying to find a parameter set, where the model exhibits a 3-Spike-Burst.
Therefore I changed the conductances on the biophysics page in the cell builder. Finally I found a a set of parameters showing the desired behaviour. I saved the session and closed NEURON. Afterwards I started NEURON again and loaded the saved session. Surprisingly the model exhibited only a 2-Spike-Burst. Checking every parameter three or four times I'm sure that the parameter sets are identical. Then I changed one of the conductances in the cell builder, but reversed it immediately. I ran the simulation again and the model showed the desired 3-Spike-Burst.

So in a nutshell I did:

- found parameter set, which produces a 3-Spike-Burst
- saved session
- closed NEURON
- started NEURON, loaded session
- ran simulation -> 2 Spike Burst
- changed arbitrarily one conductance and reversed the change immediately in the cell builder ***
- ran simulation -> 3 Spike Burst

*** Maybe it is useful to mention that changing the parameter in hoc (with a script) does not lead to a 3-Spike-Burst, but to a 2-Spike-Burst

I use Ubuntu and NEURON 5.9, but I tried it also with NEURON 6.0 on a windows XP-machine.

Does anyone have an idea what the cell builder does, when a parameter is changed? "Continous Create" is on.

By the way I exported the cell to hoc and tried to reload it, but it produces an error

procedure too big

. The model is pretty large (256 sections with different parameters). Can this be connected to my problem ???


Thank you

Marie-Therese

[ted]

There's a bug, but the question is where. To find out, we'll have to be able to reproduce the
problem. Can you zip up all the hoc, ses, and mod files that are involved, and send them to
ted dot carnevale at yale dot edu?

Does anyone have an idea what the cell builder does, when a parameter is changed? "Continous Create" is on.

With Continuous Create on, changing any biophysical parameter* other than Ra or Cm will
merely cause an immediate update of the corresponding coefficients in the system
equations, without affecting the number of ODEs that describe the system. Changing a
section's Ra or Cm will affect the related coefficients; if you're using the d_lambda rule, it
may also force a change in nseg, thereby altering the number of system equations.

*--biophysical parameters are all parameters other than L and diam (which are geometry
parameters) and the spatial discretization parameter nseg.

I exported the cell to hoc and tried to reload it, but it produces an error
procedure too big

Has nothing to do with the problem.

[horstmann]

Thank you for your help and your quick answer.

With Continuous Create on, changing any biophysical parameter* other than Ra or Cm will
merely cause an immediate update of the corresponding coefficients in the system
equations, without affecting the number of ODEs that describe the system. Changing a
section's Ra or Cm will affect the related coefficients; if you're using the d_lambda rule, it
may also force a change in nseg, thereby altering the number of system equations.

Thank you for the description. So maybe changing a parameter and reversing it, could force a change in nseg and therefore alter the spatial accuracy. Could be... But it is no good news if the model is so highly dependent on numerical accuracy....

I send you everything you requested, don't be shocked it's a very large model with some slow time constants. :-)

exported the cell to hoc and tried to reload it, but it produces an error
procedure too big

Has nothing to do with the problem.

Ok, I thought so. I just wanted to check if the problem also occurs, if everything is written in hoc code to avoid possible bugs in the gui. Unfortunately I did not manage to get the hoc code run (although I solved the problems with too large procedures).

Thank you for your quick reply

Marie-Therese

Edit: The same phenomenen occurs if I don't change an arbitrary parameter, it is sufficient to click on parameters (on the biophysics page of the cell builder)

[ted]

So maybe changing a parameter and reversing it, could force a change in nseg and therefore alter the spatial accuracy.

Only if the parameter was Ra or Cm and only if restoring the parameter failed to also
restore nseg (which would mean there's a bug in the Cell Builder's "continuous create"
function).

[horstmann]

Ok, than I can exclude that. I didn't change Ra or Cm, so nseg shouldn't have changed.

[hines]

Sorry about that. On loading the
session file, not enough was done to
establish the consistence between
geometry and biophysics. The small change indicated at
http://www.neuron.yale.edu/cgi-bin/trac ... geset/1955
should fix things.
With regard to the "procedure too big" problem with exported hoc files, I went
through the print routines and broke the generated procedures into acceptable size pieces. That code change is not something you would want to do manually but you can use
http://www.neuron.yale.edu/cgi-bin/trac ... geset/1956
and go to the bottom of the page to
generate a unified diff file and apply it
with patch. (I assume you are working
on unix) Or just get the 5 indicated files and replace your version. I should
mention, though, that one of the distal styles you are using (normalize so
all distal ends at 1) is not implemented
for exported hoc files. So if you need that style, you must use the CellBuild
to define the cell. And you may want
to do some plots of your parameters to make sure that is what you think it should be (ModelView is good for that)

[horstmann]

It works!!!

I'm really happy about it. Thank you for your work and the really quick replies!

Marie-Therese

[horstmann]

There occurs a similar error, when changing a parameter in the parameterized section list (the _x list). You have to set (and immediately reset) another parameter in a not parameterized section list or a section, e.g. in the soma to apply the changes, otherwise it seems that the parameters of the dendritic tree (which has parameterized mechanisms) are not correctly initialized. Maybe you can fix that, too???

Thank you

Marie-Therese