[gmx-users] Electric field or CompEl protocol?

alex rayevsky rayevsky85 at gmail.com
Sun May 20 15:44:44 CEST 2018

)))) If I find resources for several calculations (the last one continued
for 20 days) I'll try both methods and will share You. It seems I should
find another smaller test system and reproduce the approaches.
Thank You

Sun May 20 09:55:14 CEST 2018

I think CompEl is described in the manual, the options for it are
here: http://manual.gromacs.org/documentation/2018/user-guide/mdp-options.html
I've never used it, so I can't suggest anything, but you can ask on this
board for specific mdp examples.
The way CompEl works is conceptually simple: it maintains a
transmembrane ionic concentration gradient by swapping ions across the
periodic boundary. The result is that on average you get a voltage of
(kT/q)log(c_above/c_below) across the system. In reality, you get noisy
results (see Fig. 3 (b) in http://www.mpibpc.mpg.de/grubmueller/compel
). On the other hand, if you have long simulated times, you can still
get clean data.

The problem with using a constant field is that it is only physical for
a system with a nearly constant dielectric throughout. I am guessing
that is not your case and you have water (epsilon ~80) and a lipid
membrane (eps ~ 5?). If there was a real voltage across such a box, it
would almost entirely drop across the membrane (i.e. high field across
membrane and low field elsewhere). This is why I prefer to use fields
that are as low as computationally possible.

I would try CompEl at least out of curiosity. In principle, it is a
solid idea, but I think this algorithm is clunky and how it agrees with
PBC is unclear to me. If you get clean and reasonable data, please let
everyone know! :)


On 5/20/2018 1:16 AM, alex rayevsky wrote:
>* Dear Alex!
*>>* Yes, I thought about all Your reflections and I'm also not sure that CompEl
*>* is well parameterzied for a non-specialist like me and the electric field
*>* is more intuitive for me. However, when I saw the dimension 'V/nm' for the
*>* first time, I thought that something must depend on the length of the  axis
*>*    of application (in my case it is about 12 nm) or the thickness of the
*>* membrane.
*>* These two of 20 articles I've found on the theme befor wrote in gmx
*>* society:
*>* Structural and Functional Effect of an Oscillating Electric Field on the
*>* Dopamine-D3 Receptor: A Molecular Dynamics Simulation Study. ( DOI:
*>* 10.1371/journal.pone.0166412 )  and Molecular dynamics of ion transport
*>* through the open conformation of a bacterial voltage-gated sodium channel.
*>*   ( https://doi.org/10.1073/pnas.1214667110
<https://doi.org/10.1073/pnas.1214667110>). ANd this method works
fine, in
*>* general, they've got what they wanted.
*>* But there is no full description of parameterization. what can You say?
*>>* Thank You
*>>>* At the same time compel method is very popular too, here is
another mention
*>* of CompEl - http://dx.doi.org/10.1016/j.bpj.2017.02.016
*>>>* Alex <https://www.mail-archive.com/search?l=gromacs.org_gmx-users@maillist.sys.kth.se&q=from:%22Alex%22
*>* Sat, 19 May 2018 17:35:15 -0700
*>* <https://www.mail-archive.com/search?l=gromacs.org_gmx-users@maillist.sys.kth.se&q=date:20180519
*>>* It's more of a philosophical question in, unfortunately. I don't use
*>* CompEl, because I believe it is conceptually clunky, but that's a
*>* matter of opinion that could turn into discussion beyond the scope of
*>* your question. I don't study biomolecules, so I can get away with
*>* applying direct fields. For biomolecules, however, I do suggest at
*>* least looking into CompEl and how it works, and then choosing
*>* appropriate setup sothat you do not slow down your simulation too
*>* much.
*>>* That said, 0.4 V/nm does not really correspond to 40 mV in any
way. The best
*>* "fake" guess is that the voltage drop across the entire box is its height,
*>* times the value of E-z. It is fake, because your field has nothing to do
*>* with the solution of the Poisson's equation, or the box height. The
*>* consequences of this field do, but the field itself doesn't, if that makes
*>* sense. One other point to be made: water's dielectric breakdown threshold
*>* is around 100 MV/m = 0.1 V/nm. Noone in the community that publishes in
*>* Biophysical Journal seems to care about it, but huge simulated fields can
*>* be incompatible with what's being studied.
*>>* My response probably doesn't help much, but this is the situation with all
*>* MD software that relies on Ewald summation.
*>>* Alex
*>>>* On 5/19/2018 5:16 PM, alex rayevsky wrote:
*>>* Dear all,
*>>* Which protocol, Electric field section or the CompEl, I should use in the
*>* situtation:
*>* 1. I built an ion channel by homology, prepared a bilayer membrane, embeded
*>* my protein and run a simulation to relax the system (100 ns)
*>* 2. my channel was closed all the time.
*>* 3. I want to run four parallel simmulations, starting from the relaxed
*>* state:
*>* a) system under the effect of -80 mV and under +40 mV - the second one
*>* should cause a pore opening;
*>* b) both previous variants with a ligand in the pore;
*>>* The voltage sensitive domain of the Nav channel should respond to the
*>* electric stimuli, that is why I thought it is reasonable to apply it to Z
*>* direction and assign electric-field-z = 0.4 0 0 0 for +40mV state, for
*>* example. other parameters should stay intact, I think, because I don't know
*>* if they should be changed...
*>>* at the same time I've read several different works when CompEl was
*>* implemented to the membrane-channel systems. The end of the
*>* pagehttp://www.mpibpc.mpg.de/grubmueller/compel duplicates a gromacs
*>* manual,
*>* however I didn't find any mention of a voltage handling and what exactly
*>* I'll obtain at the end....
*>* Which method is more approrpiate for my task?
*>>* Thank You !!
*>>>>>>* *Nemo me impune lacessit*

*Nemo me impune lacessit*

More information about the gromacs.org_gmx-users mailing list