opensourcebrain/korngreen-pyramidal

porting Alon Korngreen's pyramidal neuron model to neuroml

L5 Pyramidal Neuron - Almog and Korngreen 2014

NEURON and neuroConstruct versions of Alon Korngreen's pyramidal neuron model

See http://www.opensourcebrain.org/projects/almog-korngreen-pyramidal-neuron

korngreen-pyramidal's Issues

incorrect parameter name in original model

In the original (modeldb) file params.hoc, line 64:
ca_qh=2

should instead read
cah_qh=2

This misnaming is innocuous, since the qh par is set to 2 in ca_h.mod and is not changed elsewhere.

parameter names

the original hoc files make heavy use of parameters in the mod files. Even if these parameters are defined in cml as <parameter>, the translation to .mod will include prefixes that make them impossible to be accessed from the original .hoc files. This is particularly relevant for the "shifts" in cah.

There is a lot of redundancy in the generated mod files (possibly introduced by the automatic conversion from cml to neuroml2, see duplicate Constants in all comp types) in parametrized values, which can contribute to #3

hardcoded 1e-4 in gbars

The original .mod files have a hardcoded *(1e-4) scaling factor to the maximum conductance. What would be the best way to capture that in the cml transcriptions?

q10 parameter in iH

q10 should be a parameter in iH, not hardcoded (it is changed from the hoc). As of 1.8.1 channelml does not allow using a parameter for the q10_settings q10_factor attribute.

mod generated from cml make simulations much slower

When using the .mod generated from cml (after replacing the appropriate parameter gmax->gbar permeability->pbar and shifts) in the original model, the simulations run much slower. Evidently, the .mod from cml have much more RANGE and ASSIGNED stuf, as well as intermediate vars, but I would expect those to be optimised off after compilation.

q10 temperature corrections - kslow

For kslow, there is a q10 correction for a kinetics, but none for b, b1.
Also, there seems to be missing parentheses for the correction in beta , a6 being left out.

scopmath error when using ca_h from cml with fixed stepsize

After replacing the relevant parameters in the mod file with shift/shifth (which are changed by the hocs in the original model), it is only possible to run the original model with variable stepsize integration - otherwise, a scopmath error is raised on initialization. I suspect that can be related to #3

q10 temperature correction in SK

In the original modfile SK.mod, tau is incorrectly multiplied by the q10 factor, instead of divided. That traces back to a previous version of the mechanism from Deister et al 2009

I have contacted Alon Korngreen and he is now aware of this issue.

The NeuroML version of SK implements the correct q10 scaling.

qh_cah set but not used

in the original model, in params.hoc, line 63:

cah_qm =2

is set but never added to any section, i.e., this par gets its value from the modfile (qm=4).

unused pars in the original model

in the original (modelDB) model, at model.hoc
vshift_ca
vshift_na

are defined but not used by the mechanisms.

kslow kinetics - fractional conductance

When translating kslow to cml, my first natural choice was to have three gates a, b, b1 (in accordance to what is expressed in the original mod file). However, the generated mod was calculating the fraction as (0.5b * 0.5b1) instead of the expected (0.5b + 0.5b1). By making b1 into a second open state for a single gate b, the fractional conductance is generated correctly. Is that the expected behaviour in nml? Is it documented?