Reproducing

[AlexanderMath]

Question 1: Is the hard knock-on expected at exactly 0.1eV/A? (e.g. see below from paper)

Details:
Ran python KcsA.py, but changed F_magnitude = 0.05 to F_magnitude = 0.10 to match the paper:

The potassium ions skip the water molecules (i.e. a "hard knock-on').

[AlexanderMath]

Ran F_mangitude=0.05 for ~40ps - the potassium ion also skips the water molecule.

Question 2: Would it be cheating to move the water molecule that gets skipped a little further down the channel? (it seems to be much closer to the potassium ion than the other water molecules)

[AlexanderMath]

Video https://www.youtube.com/watch?v=8_Udxymd8vo (should also be able to see trajectory if you click "Dynamics" on rcsb.ai)

[timduignan]

Hi Alex, thanks for this! Yes sorry that is not the best starting structure as it hasn't been pre-equilibrated, i.e., run without the external force on the potassium. This is best practice so I should have done it and I did see hard knock on in one of the three simulations I ran with that structure. In fact if you look at it you will see the potassium is sitting right at the entrance to the SF with no waters solvating the threonines at the entrance so I think the potassium can just jump in quickly. That is why the next four simulations where I saw the soft knock on I only applied the force after I had pre equilibrated for 60 ps to get a more physically reasonable starting structure. I have updated the repo with that better starting structure, so I would try again with that or you could run your own pre-equilibration by turning the force off for the initial part of the simulation. Otherwise I am happy to share the trajectories if you're interested, they're just too big for GitHub. Ideally, I would go to an even lower force as even the 0.05 is still much larger than physiological conditions and so will still produce more hard knock ons then in the real system but as I mention in the paper there's some instabilities of the surrounding protein at larger timescales which make that hard. I would like to get back to it and get some proper statistics on it at some point too.

[AlexanderMath]

starting structure as it hasn't been pre-equilibrated

Ahh, ofcause, missed that, thanks!

I have updated the repo with that better starting structure,

Awesome! 🚀 Will kick it off Tuesday evening.

Otherwise I am happy to share the trajectories if you're interested, they're just too big for GitHub.

That'd be neat! Perhaps Google Drive / Dropbox would work?

Alternatively, would it be annoying to use smth like

!pip install logmd
import logmd
dyn.attach(logmd)

Then gives link to rcsb.ai/run=1231e1 with webviewer + download button? Or maybe we'd want logmd prev_run.traj to circumvent running it again..

Ideally, I would go to an even lower force as even the 0.05 is still much larger than physiological conditions and so will still produce

Yes - actually have a run going with force=0.025 (will try force=0.005 at some point; would also be cool to try disease mutated version).

[timduignan]

Great I'll send you a dropbox link once the uploads ready. An interactive web viewer would be awesome actually, as long as it's not too slow with the big trajectories.

Great let me know how the lower forces look, I had one go but the entrance seemed to dehydrate because it was too slow to conduct. The problem seemed to be the histidines below the SF deprotonate and the protein seems to try to change conformation, which is the right thing to do as its meant to close at neutral pH but it can't really do it properly as I put the constraints on the outer part, so I really just need to simulate the whole protein in the membrane ideally.

Another interesting test I haven't tried is running pure water through it as this channel is known to be a good water conductor so I suspect you might see additional flipped carbonyl groups to enable the fast water flow.