opensourcebrain/vervaekeetal-golgicellnetwork

GitHub repository for an OSB project containing a cerebellar Golgi cell based on Vervaeke et al 2010

vervaekeetal-golgicellnetwork's Issues

Numerical issues with NaT_CML.nml channel

Hi @pgleeson,

Do you know why the nml2 version of the NaT channel would have numerical issues?

I have a simple voltage clamp protocol that runs on the order of minutes and results in a few hundred MB .dat files with many other channels.

However, this channel+protocol combination blows up NEURON even with dts as small as 0.00001. When I modify the resulting neuron nml_.py file to run using cvode, the simulation takes hours and the resulting vectors or dat files crash 80GB ram machine or run out of disk space.

Any ideas what might be happening with this channel?

The attached zip contains a self-contained LEMS file with the protocol (+channel), and can be run with jnml -neuron

LEMS.zip

KBK/KC issue

There was a change in the mod file used in the original https://github.com/OpenSourceBrain/SolinasEtAl-GolgiCell model in Neuron in the https://github.com/OpenSourceBrain/SolinasEtAl-GolgiCell/blob/master/NEURON/Golgi_BK.mod channel.

Originally a table was used to calculate the rates:

PROCEDURE rate(v (mV)) {LOCAL a_c, b_c 
	TABLE c_inf, tau_c 
	DEPEND Aalpha_c, Balpha_c, Kalpha_c, 
	       Abeta_c, Bbeta_c, Kbeta_c, celsius FROM -100 TO 30 WITH 13000 
	a_c = alp_c(v)  
	b_c = bet_c(v) 
	tau_c = 1/(a_c + b_c) 
	c_inf = a_c/(a_c + b_c) 
}

However, this was only using the value of [Ca2+] at t=0 in this Ca2+ dependent channel, whereas it should have used the instantaneous value of v and [Ca2+].

This issue was fixed on ModelDB (the TABLE statement removed) and in the current Solinas et al impl here, and the NeuroML 2 impl of this reproduces this behaviour: https://github.com/OpenSourceBrain/SolinasEtAl-GolgiCell/blob/master/NeuroML2/Golgi_KC.channel.nml.

The behavior of the channel was slightly different, and the conductance density of this chan in the solinas cell model was changed to compensate and reproduce the original behaviour

However the original issue propagated to the detailed cell model in this project. This mod file channel was used in the nC version in the 2010 paper, as opposed to the corrected version.

The NeuroML implementation of this cell needs to take this into account, along with other issues such as OpenSourceBrain/SolinasEtAl-GolgiCell#4 and #1.