NEURON vs GENESIS - Opinions, please?
Matt Jones
Has anybody out there used -both- NEURON -and- GENESIS, and would like
to comment about which is a better choice for detailed compartmental
modeling of synapses, spiking and networks (on Mac computers)?
Some issues:
1) I intend to use Macintosh computers -only-, so does that preclude
using GENESIS (I could possibly compile it in unix for OS X under
FreeBSD).
2) I want to be able to use synapstic receptors with specific kinetic
schemes, not just alpha functions.
3) I'll be simulating lots and lots of synaptic inputs, and they all
need to be independently controlable.
4) I want my initial single neuron simulations to be easily scalable
up to many many neurons in a network, without a lot of rewriting of
model files.
5) I'd like to be as flexible as possible with respect to job control.
The ability to drive lots of simulations from a script or batch file
is important.
Any opinions?
Thanks,
Matt
Matt Jones
> Hi all,
>
> Has anybody out there used -both- NEURON -and- GENESIS, and would like
> to comment about which is a better choice for detailed compartmental
> modeling of synapses, spiking and networks (on Mac computers)?
>
>
>
> Thanks,
>
> Matt
Not that anyone cares, but I've decided on GENESIS.
mat
I was also wondering whether to opt for genesis or neuron. Certainly
Neuron seems to have been used in more publications, but almost all
seem to be studies of single neurones rather than networks which I
would be more interested in. As a side-project I'm currently learning
Fortran and (attempting) to program my own simulator, which hopefully
should give me much greater control and also slim down the software so
that it runs faster.
Why did you decide on Genesis?
How have your initial 'experiments' gone?
Cheers
Richard S. Norman
An awful lot of work has gone into developing projects
like Neuron and Genesis. It might be fun to try to do your
own simulator, but you will spend an incredible amount of
time before you even reach the level of these programs,
let alone surpass them.
So build your own for fun or as a learning tool. Don't do it
if you have "real" work to do!
Matt Jones
Wow, someone did actually care....
Here's an interesting piece of trivia:
Question: What is the most popular compartmental modeling software in
the literature?
Answer: "Other" - (i.e., neither Neuron nor Genesis, but usually a
custom-written program for the purpose of running just that one
simulation)
I decided on Genesis for 2 reasons:
1) There is actually documentation available for Genesis, and a fairly
active user maillist, an actual textbook that guides you through its
use, etc. These exist for Neuron too. The user maillist is probably
even more active. But the "documentation" available for Neuron is
pretty disorganized and sparse. Also, my experience with Neuron is
that the user interface is pretty clunky, and in a previous
incarnation of Neuron (for Mac) I found it impossible to properly save
a simulation and re-run it later with the same settings. I'm sure
that's been fixed by now, but it was enough to put me off. Don't know
about Genesis' interface yet.
2) Genesis is built to work like unix, which almost certainly means it
is extremely flexible and expandable (and indeed, the designers had
large scale network simulations in mind). I've been using other large
scale unix simulation software recently, and have found it incredibly
useful to be able to write a script (in Matlab) that writes a script
(in unix) that runs hundreds of separate simulations and organizes the
output, and then pipes it into an analysis script, etc. I literally
spend the afternoon setting up the scripts, then run them just before
I go home, and when I come in the next day the simulations have been
run, and the data have been analyzed according to my specifications
and the figures have been made (!). Well, ok, it's not quite as simple
as that. But working in unix has a lot of advantages for automating
various things.
Haven't actually run any Genesis simulations yet, but have ordered the
book and downloaded the binaries (precompiled for Mac OS X no less -
how perfect is that?) and demos.
Before you launch into coding your own simulator, I would recommend
having a look at the Genesis or Neuron code itself, because there are
apparently a number of useful tricks for making solving all those
differential equations run much more efficiently. At least this is
what I've been told.
Cheers,
Matt
mat
neuron and genesis by many parties, but as Matt noted, most papers
utilise custom written software that only runs that specific
simulation and small variations thereon, rather than the catchall of
neuron/genesis (e.g. Traub et al., Edelman). Neuron and Genesis are
undeniably huge and complex, but how much of the program does any one
user make use of?
Also I think it would be useful to try and do the coding, may even
give me greater understanding (!)
Was thinking of using Matlab to analyse the data from the simulator
(another way of saving on the amount I have to write compared to
Neuron/Genesis which include custom analysis/graphic tools), is it
relatively simple to integrate with other programs does anyone know?
I've read of the ability to compile Matlab analysis and graphing
routines into standalone programs, has anyone any experience of this?
Christian Wilms
> Haven't actually run any Genesis simulations yet, but have ordered the
> book and downloaded the binaries (precompiled for Mac OS X no less -
> how perfect is that?) and demos.
I would be very intrested in hearing how your first steps with Genesis
on MacOS X were. Does it run without requiring an X-server?
I just decided, that I woul prfer going with Neuron, because it runs on
more platforms and I like the "section"-concept for modelling neurons,
compared to directly modelling single compartments in Genesis (or did I
miss an important development here?). Does Genesis/MacOS X come with a
running version of GMODL?
thanks for any answers, Chris
Matt Jones
>
> Also I think it would be useful to try and do the coding, may even
> give me greater understanding (!)
>
I agree 100%. I learned an awful lot by writing my own simulator for a
small network of 3 HH neurons coupled by dynamic synapses (in Matlab).
The problem came when I wanted to -add- a neuron. I essentially had to
rewrite the code from scratch. This is probably the -main- advantage
of using a canned program. They've already implemented a built in
"scalability" (especially Genesis).
On the other hand, figuring out your own solution to the scalability
problem would sure teach you a lot too.
> Was thinking of using Matlab to analyse the data from the simulator
> (another way of saving on the amount I have to write compared to
> Neuron/Genesis which include custom analysis/graphic tools), is it
> relatively simple to integrate with other programs does anyone know?
Yeah, Matlab is great. I use it for everything. It's very easy to
import data in text format, or standard image formats. If you need to
analyze something in a custom binary format, that's not too hard
either as long as you know the format exactly. I haven't -actually-
used it to analyze GENESIS output yet, but I would be shocked if there
was any problem doing that.
> I've read of the ability to compile Matlab analysis and graphing
> routines into standalone programs, has anyone any experience of this?
I've never done it, but i know there are some limitations. I don't
think the Matlab compiler handles the struct data type properly, and
there may be some issues with graphics too (because a lot of Matlab
graphics is object oriented, and probably relies on structs.
Another option with Matlab is that you can compile your own code, then
call that with Matlab. I find Matlab really great for writing GUIs, so
you could have a Matlab GUI front end for your simulator, the
simulator as compiled code, and a GUI backend for analyzing the
output. That would be efficient, because solving the diffeqs will be
the really cpu-intensive part so you would want that compiled. But you
may not want to write a whole GUI in fortran (can fortran even do a
gui?).
Oooh. This sounds like fun...
Cheers,
Matt
mat
> rewrite the code from scratch. This is probably the -main- advantage
> of using a canned program. They've already implemented a built in
> "scalability" (especially Genesis).
>
> On the other hand, figuring out your own solution to the scalability
> problem would sure teach you a lot too.
:) As always with these things I began thinking that it would be
reasonably straightforward affair to encode the number and branching
structure of neurons along with relevant global and compartmental
parameters in a few matrices which I would then run through a
integration algorithm. I had not really considered computational
efficieny at the outset but this seems to be a 'big thing' to put it
mildly. Looking through some of the documentation for the
aforementioned simulators it seems that encoding the dendtritic
arborisation in matrices and then reducing this to a tridiagonal
matrix is at the very heart of both neuron and genesis, but I'm
unclear as to how exactly the structure is "encoded" in the matrix.
I've been looking for the original paper by Hines (1984) but it seems
to have been published in a rather obscure journal.
Can anyone point me to other sources explaining the Hines matrices
etc. The way I was thinking of doing it was a much more explicit
almost compartment-object approach, but this intuitively seems to
involve a much greater cpu overhead. Also, does anyone know of good
sites dealing with the finer points of numerical integration
techniques, especially implicit.
>
> Another option with Matlab is that you can compile your own code, then
> call that with Matlab. I find Matlab really great for writing GUIs, so
> you could have a Matlab GUI front end for your simulator, the
> simulator as compiled code, and a GUI backend for analyzing the
> output. That would be efficient, because solving the diffeqs will be
> the really cpu-intensive part so you would want that compiled. But you
> may not want to write a whole GUI in fortran (can fortran even do a
> gui?).
This was essentially my idea too, to write the minimal amount of code
required for the core integration routines and then analyse the data
in something like Matlab. You can buy third-party routines to create
fortran GUIs (e.g. GINOMENU), but I personally do not really want to
spend much time doing that as I reckon a simple command-line interface
is just as powerful and probably easier to implement, though I have no
experience of creating GUIs in Matlab.
>
>
> Oooh. This sounds like fun...
Should be, hopefully.
Matt Jones
> Matt Jones <jone...@physiology.wisc.edu> wrote:
>
> I would be very intrested in hearing how your first steps with Genesis
> on MacOS X were. Does it run without requiring an X-server?
I'll post my progress here.
I'm not sure Genesis -requires- an X-server (does it?), but they
recommend using XFree86/XDarwin, which is available from
http://sourceforge.net/projects/xonx/.
>
> I just decided, that I woul prfer going with Neuron, because it runs on
> more platforms and I like the "section"-concept for modelling neurons,
> compared to directly modelling single compartments in Genesis (or did I
> miss an important development here?).
I am not sure, but somehow I would be surprised if you couldn't assign
sections in Genesis and manipulate them as necessary. I'll let you
know what I find out.
>Does Genesis/MacOS X come with a
> running version of GMODL?
Again, not sure yet, but my understanding is that the OSX version is
complete.
Cheers,
Matt