The Traub'94 simulation has been implemented by Pulin Sampat
(Brandeis University) and by Patricio Huerta (MIT) with
help from Dr. Roger Traub.
The Traub91 simulation has been modified to incorporate the new
model by Traub, et. al. [Journal of Physiology, Vol. 481.1, p. 79
(1994)]. Basically, the new model should replicate (almost exactly)
the '91 results. Some of the new features of the model are:
1. Implementation of dendritic branching. The model now has
sixty-four soma-dendritic compartments.
2. Implementation of an axon initial segment, and an axon compartment.
3. Model is constrained using antidromic stimulation.
4. Implementation of spines on dendritic compartments. This was
done by doubling the area (in his paper). However, to incorporate
this in GENESIS, I had to double CM, halve RM, and double the
ionic conductance densities for each compartment.
5. I have not yet implemented synaptic channels.
6. Time step is 2.5 microseconds; much smaller than the 1991 paper.
So the simulation is a lot slower, when used with non-hsolve methods.
There are other subtle changes; please refer to the paper for the
differences.
=========================================================================
Additional comments by Dave Beeman (dbeeman@dogstar.colorado.edu)
As this model has a large number of compartments, it should be run using
one of the implicit integration methods (method 10 or 11), using hsolve.
This will allow a step size of about 25 microseconds, instead of the
2.5 microsecond step required for the default method 0. This is illustrated
in the script "traub94.g", which uses hsolve with chanmode 1, and method 11
(Crank-Nicholson). The "userprefs.g" file included here allows it to
be used with Neurokit.