axon_model.m

models.axon_model

models.axon_model is the 'core' of the NEURON-MATLAB interface implemented in ViNERS. models.axon_model generates and runs multithreaded NEURON simulations, executing tools.make_from_template to generate customised axon models with specified ultrastructural and biophysical properties. The expectation is that the model is called in a parfor loop like so:

parfor ii = 1:nSimulations 
    result(ii,:) = axon_model(ii, VE_field or [], model, arguments{:});
end

where VE_field is the output of tools.load_Ve_field.

models.axon_model is called by models.membrane_currents, models.axon_thresholds, and models.nerve_stimulation, and utilises models/Gaines2016_v3 and models/Gaines2016_v3. See also models.analysis_interface.hoc

Syntax:

• Sundt, Gaines, MRG : select model to simulate. models/Gaines2016_v3 contains the implementations of both the Gaines (myelinated afferent) and MRG (myelinated efferent) models. models/sundt2015_v1 contains the implementation of the Sundt model.
• -debug, -echo : Create a duplicate NEURON process in a shell so that the execution of the NEURON code can be debugged
• -stimulus : set stimulus waveform and possibly current levels. See plots.preview_stimulus for more details. if -stim is not set, construct a biphasic stimulus using -pw [pulse-width (ms)] (default: 400 µs, with a 20 µs inter-phase gap) and -delay [time (ms)] (default: 30 ms). for -fixed-stim the parameter -uA is also relevent as it sets stim levels (default: 10 µA - 1 mA, 25 steps).
• -fixed-stimuli : skip threshold-finding procedure and use the values in stimulus.a (µA) to compute axon responses to predeterimined stimulus levels. Used by models.nerve_stimulation.
• -sweep-range [init LB UB] : set initial guess and lower/upper bounds for threshold-finding procedure. Ignored if -fixed-stim is set.
• -constantlength [val (mm)] : set length of simulated axon. If not set, the unmyelinated axon has a standard length of 2 mm and the myelinated axon length depends on the fascicle diameter (31 nodes are simulated). Most of the time I set this to 8 mm for myelinated axons and 6 mm for unmyelinated axons.
• fibreDiam [diameter (µm)] sets the outer (fibre) diameter.
• g_ratio [g], -gratio [g] sets the g-ratio, the ratio between the inner (axon) diameter and the outer (fibre) diameter. This is used to generate models which match the distributions generated using models.axon_population
• -xy [intercept or 3d trajectory (mm)] : set axon location

Model output flags:

• -spike : return output spiketimes, computed in neuron
• -volt : return membrane voltages at all nodes (slower)
• -current : return membrane voltages and currents at all sections (nodes and internodes). This gets bottlenecked by disk IO and is very slow, do not do this unless necessary.
• -spk-volt [AP_threshold (mV)], default -20 mV. What Vm is considered the rising edge of a spike?
• -t-start [time (ms)], default 0 ms. Don't start writing output until t >= t_begin, helps make smaller files and faster simulations.

Other arguments:

• -sphase [value or 'rand'] : set spatial phase of the axon (displace along trajectory by ±1 node length)
• simDuration [time (ms)], default 100 ms. This is probably overkill. The duration of recordings generated using the outputs of models.axon_model does not depend on this setting.
• -get-parameters : return list of axon model parameters (matches inputs needed by tools.make_from_template for models/Gaines2016_v3 or models/sundt2015_v1. Each item in this list of parameters is also an acceptable 'parameter-name',[value] input.
• -sensitivity [values] : argument vector for Sobol' sensitivity analysis.
• -no-analysis : skip analysis of NEURON output to extract parameters such as threshold and conduction velocity.
• -vClamp : Run 3-level voltage-clamp-type experiment. -inactivation runs an inactivation experiment (different set of stim amplitudes and timings). -vClamp-full returns the currents and voltages for each stimulus level (at the voltage-clamped node, settable using -VClamp-node [node_id])
• -postprocess [handle] : run a custom post-processing step instead of the standard make_cache_file. handle must be of the form [waves,result] = handle(cache_file,waves,vars,result)
• -term : instead of sampling the middle of the axon, sample the distal end of the axon (returns a slightly different current profile).
• -r-vector : return result as a vector not a struct

Operational notes:

On Windows, if NEURON runs into an error it will stop and wait for user input. With models.axon_model, this manifests as the program hanging. Thus, if models.membrane_currents or models.axon_thresholds is just sitting there doing nothing for a very long time (more than a few minutes elapse between pressing 'run' and the appearance of the first [002] Spike at: 0.11 uA or similar), try cancelling (ctrl-C) and running models.membrane_currents('-debug','-no-parallel'). This will show the NEURON console in a duplicate process, allowing you to see what error has occurred. The most common source of error is due to nrnmech.dll was not correctly copied from ~/source/nrnmech/. This will be the case if the console window is showing you something like:

        1
Reading stimulus files
begin model setup
        1
C:/nrn/bin/nrniv.exe: node_sensory is not a MECHANISM
 in C:\Users\ceiber\AppData\Local\Temp\pn-mdl-6244\/nrn-00.hoc near line 141
                insert node_sensory // mod file declaring nodal channel properties

This should look like:

       1
Reading stimulus files
begin model setup
        1
from ../models import analysis
        1
Doing internal threshold
Writing file C:/Users/ceiber/AppData/Local/Temp/pn-mdl-6244//NEURON_vm0.dat
Changed dt

Another error I experienced recently (which was diagnosed using models.membrane_currents(opts,'-debug','-no-parallel')

Can't load NEURON resources from /cygdrive/C/nrn/lib/nrn.def\[aults\] 1 
Reading stimulus files
begin model setup 
C:/nrn/bin/nrniv.exe: Couldn't find: noload.hoc 
in C:\\Users\\ceiber\\AppData\\Local\\Temp\\pn-mdl-3948/nrn-00.hoc near line 37 load_file("noload.hoc")

This was a NEURON installation issue with NEURON 8.0, uninstalling and reinstalling NEURON with the default settings fixed the issue.