Hi everyone, has anyone successfully added a gap junction to an existing model using NESTML? I would like to add gap junction support to the aeif_cond_alpha neuron model but it is unclear how this can be done within NESTML. As suggested by Charl, I started to look into updating the jinja templates but I am not sure this will be possible since both the input/output of the neuron and the model's equation must be updated to support gap junctions. If anyone has any suggestions or if anyone has NESTML code for adding a gap junction to an existing model, I would be very appreciative. Thanks!
Hi, thanks very much for your inquiry!
To add a gap junction to a NESTML model, a few modifications to the toolchain would be required. A first idea would be to add a new input port with a type of "gap". If a gap junction port is present, we could then generate the handler for a GapJunctionEvent, as well as to generate the necessary code for the waveform relaxation mechanism in the update() method of the neuron.
If you'd like to do a deeper dive into this together, I would be happy to have a video call this week or the next (after that I will be in Marseille for the HBP Summit). Please let me know what options would work best for you.
Yours, Charl
On Tue, Mar 14, 2023, at 11:10, bstrohmer@gmail.com wrote:
Hi Charl, Yes, I would greatly appreciate a deeper dive into this with you. Can you write to me directly at bstrohmer@gmail.com so we can find a suitable time? Thanks so much!
Best, Beck
On Wed, Mar 15, 2023 at 3:07 PM Charl Linssen nest-users@turingbirds.com wrote:
Hello Charl and Beck,
I would be interested to join the discussion, as I am currently trying to implement gap junctions too in an AdEx model with NESTML. Since maybe I'm not the only one interested, could you post the meeting time and the zoom details?
Best regards, Razvan
On Thu, 16 Mar 2023 at 01:28, Beck Strohmer bstrohmer@gmail.com wrote:
Dear Razvan, dear Beck,
Thanks to you both for the expression of interest. If you don't mind, shall we have a meeting with all three of us, and allow anyone to join who might also be interested in the topic?
Could you please fill in the date picker here: https://terminplaner6.dfn.de/p/c0006005efa01c5d2381ebac469d0ba4-171366
As a bit of background reading (although I'm sure you already found this), here is the paper detailing the NEST implementation of gap junctions: https://www.frontiersin.org/articles/10.3389/fninf.2015.00022/full
Yours, Charl
On Thu, Mar 16, 2023, at 02:02, Razvan Gamanut wrote:
Hello all,
Thank you for filling in the date picker. Let's meet this Wednesday (22nd) at 14:00 Jülich time.
We can meet here: https://webconf.fz-juelich.de/b/lin-dfx-nsx-j4y
Looking forward! Charl
On Thu, Mar 16, 2023, at 10:58, Charl Linssen wrote:
Thanks for confirming Charl! I look forward to meeting with all of you.
On Mon, Mar 20, 2023 at 12:21 PM Charl Linssen nest-users@turingbirds.com wrote:
Hi all,
We now have a prototype for support of gap junctions via NESTML.
The code is being tested right now on the "iaf_psc_exp" (linear dynamics) and "aeif_cond_exp" (nonlinear dynamics) neuron models and can be found at https://github.com/nest/nestml/pull/907
The pull request is still in a draft stage; some documentation still needs to be added, but most of all I would like to have more strict numerical validation of the results. If anyone could recommend some simple systems (for instance, involving only two neurons connected by one gap junction), in particular those that have a closed-form algebraic solution of the dynamics, or those that converge to a known solution under a set of representative conditions, this would be much appreciated.
Please let me know if you run into any trouble running the prototype!
With kind regards, Charl
On Tue, Mar 14, 2023, at 11:10, bstrohmer@gmail.com wrote:
Thanks Charl! I successfully added the gap junction to the aeif_cond_alpha neuron model using NESTML. A very rudimentary quantitative test shows that without the gap junction, there is a mean Vm difference over the plot of 0.223mV and with a gap junction, this is reduced to 0.043mV. You can see that there is an initial difference in Vm due to the neurons being initialized with different values, they converge quickly due to the gap connection.[image: gap_test_nestml.png]
On Tue, May 16, 2023 at 5:41 PM Charl Linssen nest-users@turingbirds.com wrote:
Hi all, I am trying to update the weight of the gap junctions implemented via NESTML in between runs of a split simulation. However, when I update the weight using "nest.SetStatus(synapse_collection,{'weight':some_weight})", I can see that the static synapse weights between the populations are also updated instead of only the gap junction weights. Is there a way to correct this so that only gap junction weights are updated? Thanks for any help!
Best, Beck
On Wed, May 24, 2023 at 4:03 PM Beck Strohmer bstrohmer@gmail.com wrote:
Hi Beck,
Happy to hear that the gap junctions are working (mostly) in order! Please apologize my late reply, I came down with a nasty bug for a while.
The trick here (hopefully) is to grab a different SynapseCollection, namely one that includes only the gap junctions and not the normal synapses (static_synapse). You can pass the "synapse_model" key to the GetConnections() call, which selects only the synapses by that name, for instance:
gap_syn = nest.GetConnections(pre_neuron, post_neuron, synapse_model="gap_junction")
I've added a new test for this in:
https://github.com/nest/nestml/pull/907/commits/cd24e9c8a16e908f73f26e035190...
Please let me know if this is what you're looking for.
With kind regards, Charl
On Tue, May 30, 2023, at 23:04, Beck Strohmer wrote:
Hi Charl, Thanks, this does indeed return only the gap junctions and is what I needed. However, the gap junction module seems to have a bug because when I call GetConnections prior to running the simulation, it hangs. Please see the screenshots attached, one only runs GetConnections and the other updates the weights prior to simulation. The weights are updated correctly, however, the simulation still hangs. You can see it only runs to 1% before hanging indefinitely.
Thanks for any help!
On Mon, Jun 5, 2023 at 2:13 PM Charl Linssen nest-users@turingbirds.com wrote:
Hmm, could this be the adaptive exponential IAF model membrane potentials going off to infinity, thereby hanging the solver (the same issue we discussed in https://github.com/nest/nest-simulator/pull/2755)? I assume you're still running the NEST Simulator version that has this fix applied. Could you send me the script so I can reproduce the error locally?
On Tue, Jun 6, 2023, at 10:20, Beck Strohmer wrote:
Correct, this is the aeif_cond_alpha that is bounded so it should not go off to infinity. In addition, I added a minimum so that it cannot go to infinity in either direction (positive or negative). I have attached the nestml neuron model file I add the gap junction to. Additionally, you can find the script that hangs in simulation after the GetConnections call.
On Tue, Jun 6, 2023 at 12:10 PM Charl Linssen nest-users@turingbirds.com wrote:
-
Beck Strohmer -
bstrohmer＠gmail.com
-
Charl Linssen -
Razvan Gamanut