This is the readme for the model files associated with the paper:

Larkum ME, Nevian T, Sandler M, Polsky A, Schiller J (2009) Synaptic
integration in tuft dendrites of layer 5 pyramidal neurons: a new 
unifying principle. Science 325:756-60

These NEURON model files were supplied by Alon Polsky.

Usage: to run the model either auto-launch from ModelDB or
download and extract this archive and then compile the mod
files under

mswin: run mknrndll and select this folder to create the nrnmech.dll
linux: run nrnivmodl in this folder
mac os x: drag and drop this folder onto mknrndll

then to start the simulation under

mswin: double click the mosinit.hoc file using windows explorer
linux: type "nrngui mosinit.hoc" in the shell prompt in this folder
mac os x: drag and drop the mosinit.hoc file onto the nrngui icon

Once the simulation is running:

To run the current injection simulations run the model adjust the
proximal or the distal IClamp processes � 0.5nA works OK for the
distal and 0.9nA for the proximal processes.  To examine the effects
of NMDA spike initiation increase the gmax variable of the
�glutamate� point process. Other variables used in
�glutamate� are:

       e-the reversal potential (should be 0)
       ntar � NMDA to AMPA ratio (in this point process gmax is the
            NMDA conductance, and the AMPA conductance is calculated
            based on the ntar value. So, for example, at gmax of 1nS
            and ntar of 10, the NMDA conductance would be 1nS and AMPA
            0.1nS)
       del-onset time of the first synaptic input 
       Tspike � ISI presynaptic activations (for presynaptic input
            train) in ms.
       Nspike-number of presynaptic activations (for presynaptic input
            train). For example, presynaptic train of 5 inputs at
            100Hz can be activated with Tspike=5 Nspike=10.

Summation of different distribution of synapses can be examined with
the �Control� panel.

numS variable is the number of synapses
GmaxS is the conductance of a single synapse (similar to gmax of
�glutamate� process
Synapses can be distributed based on distance from the main
bifurcation (between SubTuftProxS and SubTuftDistS values; only when
generation is 0) or dendrite generation.
Synapses can have either AMPA and NMDA components when the (NMDA=1;
AMPA=2) value is 1 or AMPA only when the (NMDA=1; AMPA=2) value is 2.

Please note that in some cases the temperature and the GlobalRa
variables are not automatically set to the right values. Best
simulation time value is 200ms.

---
When the simulation starts it displays a default location of synapses:



and pressing "Init & Run" will produce the simulations default display:



Detailed instructions on generating panels from supplementary Fig. S2:

Here are the instructions for how to create simulations similar as to
presented at fig. S2:

For a simulation of NMDA spike (panel C) set the 'global Ra' to any
value between 80 to 90ohm-cm and press 'run'. The glutamatergic
synapse is located at the appropriate location and should generate 3
impulses with NMDA spike following the second impulse.
Setting global Ra to 85 ohm-cm produced this graph:



For a simulation of synaptic summation (panels G and H) fist set the
gmax value of the glutamatergic synapse (point process glutamate[200])
to zero. Then redistribute 100 synapses over the 3rd order branches by
setting the 'generation' value in the 'Control' panel to 3. Set the
numS to 100 (number of synapses) and GmaxS to 1 (conductance of each
synapse in nS). Hit 'run' to generate a plot similar to panel G:



In order to examine what is the effect of AMPA only synapses, change
the value of 'NMDA=1 AMPA=2' field in 'control' panel to 2. Hit 'run'
to generate the left part of plot H:



In order to increase the number of synapses set the 'numS' value to a
higher number (500 will produce a similar result as to presented at
plot H, right part):



As the synapses are distributed randomly, each run's result is unique,
and may be somewhat different from the simulation presented in the
paper.

20120326 the solve method was updated from euler to derivimplicit in
cad2.mod.  See

http://www.neuron.yale.edu/phpBB/viewtopic.php?f=28&t=592