From: Carlos Simmerling (
Date: Tue Jul 13 2021 - 16:21:28 CDT

I think doing the comparisons that you describe can be very difficult when
you change the protocol/force field at the same time as the sequence. I
would want to make sure I have both sequences set up using the same
protocols, and with the same force fields.
But yes the accuracy of salt bridge strength in various force fields is a
potential concern. I am not sure, though, that I see why extracellular salt
bridges would be different than intracellular in terms of the physics. It
could be effects from things like differences in local ions, or possibly
pH, but in principle the amino acids should have the same intrinsic
properties on both sides of the membrane.

On Tue, Jul 13, 2021 at 2:31 PM Manas Kohli <> wrote:

> Dear All,
> This question doesn't really relate to a technical issue to VMD but I was
> wondering if anyone had a similar experience and could offer some advice.
> I'm working with a membrane protein i.e. Nav1.5. Initially roughly about
> 1.5 years ago, the rat structure for Nav1.5 came out and I simulated this
> channel and a few mutants in NAMD. I used the CHARMM27 force field and
> prepared the channel in a very similar way to the potassium channel
> outlined in the NAMD membrane protein tutorial. One particular region I was
> interested in was an extracellular complex salt-bridge flanked by aromatic
> residues that formed cation-pi interactions. For the wild-type channel,
> these interactions held together quite well and they did for a positive
> control mutant as well. However, mutations in this salt-bridge disrupted
> interactions as one would expect. This was well in line with what I
> expected and quite good. However, one point of note was that I could only
> really run such a big system in NAMD for 15ns simulations or so given the
> number of mutations I had to run and the slowish speed of NAMD.
> I was interested in running these simulations for a few structures for a
> longer duration (around 100ns or so). Recently the human Nav1.5 structure
> came out (as opposed to the rat Nav1.5 that came out earlier) so I thought
> about running the simulations on this structure instead of the rat
> structure. The human Nav1.5 sequence is almost identical to the rat Nav1.5
> structure and the region I alluded to earlier was perfectly conserved
> between Human and rat. Looking at the cryo-EM structure of the Nav1.5
> confirmed this. To simulate this system for about 100ns, I decided to try
> and use gromacs as opposed to NAMD. However, the major problem I'm running
> into is that the complex extracellular salt-bridge seems to be breaking
> pretty consistently and not holding together like it did in the NAMD
> simulations. I tried multiple simulations all to the same effect:
> - I aligned my human Nav1.5 structure to a pre-existing DPPC bilayer
> removing overlapping DPPC so that the protein would sit in the membrane. I
> then solvated and ionised as normal. I performed a production simulation
> for 100ns but the complex salt-bridge split apart in this case. The
> forcefield I used in this case was gromos53a6 with lipid parameters added
> as done through Prof Lemkul's tutorial (
> <;!!DZ3fjg!paLDpnhMZlCo3lSIullup3Afln-ofrAIxKjM5DCTkig9gaLKP2YcmLms7PRvztQ4Dg$>
> )
> - I used CHARMM-GUI to prepare my system in this case. CHARMM-GUI prepared
> all the files for simulation so I just ran a production run for about 5ns.
> However, even in this production run, the complex salt-bridge broke. The
> force-field used in this case was CHARMM36
> There are a few papers that seem to suggest that extracellular
> salt-bridges are a bit finicky in MD simulations and generally hard to
> model given their interaction with ions and solvents and so it's hard to
> reliably assess their stability. One other reason this paper suggests,
> <;!!DZ3fjg!paLDpnhMZlCo3lSIullup3Afln-ofrAIxKjM5DCTkig9gaLKP2YcmLms7PQmvH9zYQ$>
> , is that CHARMM22 over-stabilises salt-bridges and that these values were
> corrected for in later versions of CHARMM which is why I think the complex
> salt-bridge may be breaking.
> I'm a bit unsure about what could be causing the salt-bridge to break in
> the more recent simulations I'm running. It could be due to the human vs
> rat Nav1.5 structure but given how similar the structures are (basically
> identical) I don't think that this is the case either. Does anyone have any
> idea what could be happening? I'd appreciate the help!
> Thanks and Best Regards,
> Manas