[neuroConstruct] New version neuroConstruct, two new NeuroML models & update to ChannelML NEURON mod file mapping

Padraig Gleeson p.gleeson at ucl.ac.uk
Wed Aug 25 16:23:41 BST 2010


A few recent developments:

1) A new version of neuroConstruct, v1.4.1, is available at 
http://www.neuroconstruct.org. This contains minor updates to the core 
application, full details are here:

One potentially useful addition in this version is the ability to load 
valid NeuroML files using the command line option -neuroml:

./nC.sh -neuroml MyNeuroML.xml

(or nC.bat -neuroml MyNeuroML.xml for Windows, see 
http://www.neuroconstruct.org/docs/install.html). This will create a new 
neuroConstruct project containing all the elements from the NeuroML 
file. For one containing a full model description of cells, channels, 
network structure etc. the neuroConstruct project is populated with all 
of the elements required for generating the model in NEURON, GENESIS, 
etc. and default values of simulation duration, dt and plots of membrane 
potential are added (including a -sedml option for loading this info 
from a SED-ML file is in the pipeline).

Some restrictions apply as usual: a NetworkML only file needs to be 
loaded (via the menu) into a project containing cell groups and network 
connections with the names of the populations/projections, etc.; 
template based NetworkML representations are not yet supported.

2) Two new projects are included with this release, and are also 
available here http://www.neuroconstruct.org/models/index.html. 
SolinasEtAl_GolgiCell is an implementation in NeuroML of the abstract 
Golgi cell model from Solinas et al. 2007: Computational reconstruction 
of pacemaking and intrinsic electroresponsiveness in cerebellar Golgi 
cells. Front Cell Neurosci.

VervaekeEtAl-GolgiCellNetwork is a project containing a number of Golgi 
cell models with detailed reconstructed morphologies, and featured in a 
recent paper in Neuron: Rapid Desynchronization of an Electrically 
Coupled Interneuron Network with Sparse Excitatory Synaptic Input: 

Both of these models can be fully expressed in NeuroML, but at this time 
only run on NEURON due to the complexity of the multistate SK channel 
(see below).

3) The XSL mapping file to generate NEURON mod files from ChannelML 
files (ChannelML_v1.8.1_NEURONmod.xsl) has been updated with the 
following two changes (note the specification for ChannelML v1.8.1 is 
unchanged, these changes just increase the scope of models that are 
supported by NEURON mod files)

  a) Mapping of multi state (more than the 1 open & 1 closed state in 
HH) kinetic scheme descriptions to mod files is now supported. 
Previously these could only be mapped to NEURON ChannelBuilder format, 
but now they can be used for generating mod files using the KINETIC 
block and statements like: ~ n0 <-> n1 (alpha_n0_n1, beta_n0_n1), e.g. 
the 5 start representation of the squid axon K channel:


is mapped to:


Another example of such a channel is the multistate 
SK/afterhyperpolarising K+ current from Solinas et al 2007, which has 
state transitions dependent on Ca2+ concentration:


This type of channel cannot (yet) be mapped to the old Channel Builder 
format, or indeed GENESIS, MOOSE or PSICS.

  b) The mapping of a decaying calcium concentration pool in an 
<ion_concentration> element is also updated. See for an example:


This mechanism now performs a check whether the section it's been placed 
on is a sphere or a cylinder: if the surface area is given by pi * diam 
* diam it assumes the section is meant to be a sphere, if not it assumes 
the section is a cylinder. This distinction is unimportant for the 
calculation of total conductance from max cond density (surf area of 
sphere of diam d is same as cylinder of length = diam = d), because if 
the model is of a pool of calcium just below the surface of thickness t, 
this pool volume is slightly different if the section is a sphere or 
cylinder. I believe the latest version covers most cases sufficiently 
well. The only problems may arise when a user wants a section with diam 
= length, wants it to be modelled as a cylinder (with no area on the 
ends) and has a relatively large thickness for the volume of calcium.


Padraig Gleeson
Room 321, Anatomy Building
Department of Neuroscience, Physiology&  Pharmacology
University College London
Gower Street
London WC1E 6BT
United Kingdom

+44 207 679 3214
p.gleeson at ucl.ac.uk

More information about the neuroConstruct mailing list