normal Regarding temperature

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago #4730 by Navarro
Regarding temperature was created by Navarro
Dear Martini users,
I want to analyze the temperature dependence of the adsorption of a Protein in a bilayer membrane. For that, I'm running simulations at 300K and 328K, but the problem is that at 300K the protein barely moves (only its lateral chains beads moves a little), while at 328K the proteins moves freely.
Is this type of comparison impossible to do with MARTINI?
This is the *mdp I'm using:
integrator = md
dt = 0.01
nsteps = 20000000
nstcomm = 10
comm-grps =

nstxout = 0
nstvout = 0
nstfout = 0
nstlog = 10000
nstenergy = 1000
nstxtcout = 10000
xtc_precision = 100
xtc-grps =
energygrps = Protein POPC DFMG W

nstlist = 10
ns_type = grid
pbc = xyz
rlist = 1.4

cutoff-scheme = group
coulombtype = Shift ;Reaction_field (for use with Verlet-pairlist) ;PME (especially with polarizable water)
rcoulomb_switch = 0.0
rcoulomb = 1.2
epsilon_r = 15 ; 2.5 (with polarizable water)
vdw_type = Shift ;cutoff (for use with Verlet-pairlist)
rvdw_switch = 0.9
rvdw = 1.2 ;1.1 (for use with Verlet-pairlist)


tcoupl = Berendsen
tc-grps = Protein POPC DFMG W_NA
tau_t = 2.0 2.0 2.0 2.0
ref_t = 300 300 300 300
Pcoupl = Berendsen
Pcoupltype = semiisotropic
tau_p = 1.0 ;parrinello-rahman is more stable with larger tau-p, DdJ, 20130422
compressibility = 3e-4 3e-4
ref_p = 1.0 1.0

gen_vel = no
gen_temp = 320
gen_seed = 473529

constraints = none
constraint_algorithm = Lincs
unconstrained_start = no
lincs_order = 4
lincs_warnangle = 30

Best,
Carlos

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4750 by floris
Replied by floris on topic Regarding temperature
Could you describe a bit better what you define by moving freely?
are you sure you did not accidentally define constraints (by putting or uncommenting: define = -DPOSRES in your mdp file)?

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago #4752 by Navarro
Replied by Navarro on topic Regarding temperature
Hi Floris,
Thanks for your reply.
What i mean is that at 300K the protein got 'stuck' in the same place during the MD simulation that i place it when i create the system with insane.
I'm completely sure that i didn't put POSRES in the *mdp, but i did use elastic network into the protein (to maintain its 'global structure' more rigid during the simulation)
This is the *itp of my protein:
Warning: Spoiler! [ Click to expand ]


One more thin that i forgot to mention before.
At the beginning of the simulation i got several warnings regarding the system pressure:
Warning: Spoiler! [ Click to expand ]

Are this warnings something to worry about?And if they are, what can i do to fix the problem?
Thanks again Floris for your help.
Best,
Carlos

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4753 by mnmelo
Replied by mnmelo on topic Regarding temperature
Hi Carlos,

The pressure fluctuations are normal in the initial equilibration steps of insane systems, which start quite far from equilibrium.

Regarding your protein freezing, and in addition to Floris' checks:

I'm assuming that other than the temperature both systems are the same.
Can you check if the water itself isn't frozen? Martini water freezes at somewhat hotter temperatures than the real thing, though most of the times one can get away with that since crystals fail to nucleate and systems remain happily fluid.

In your system you're not giving your molecules any initial temperature kick. This may help water crystals nucleate and grow on the protein or membrane surfaces, and before you know it you have a nice frozen water phase.

If freezing is indeed affecting your system it might also be useful to remove the bilayer's and the water phase's COM motion separately (read up on the 'flying ice cube effect'). Just split your system into two index groups 'Solvent' and 'Bilayer', and set 'comm-grps = Solvent Bilayer'. (Be sure, then, to put the protein in the appropriate group: 'Solvent' if it's in the aqueous phase, 'Bilayer' if it is embedded).

Cheers, and let me know.
Manel

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago - 8 years 8 months ago #4754 by Navarro
Replied by Navarro on topic Regarding temperature
Hi Manuel,
Thanks for the reply (and for the good advice!!)
Indeed the water beads were froze during the simulations (as well as some head groups of the Bilayer)!

What about if the protein start on the water surface, but i'm studying how the protein interact peripherally with respect to the headgroups of the membrane during the MD simulations?
I just run a test (100ns) using the following line in the *mdp file:
comm-grps = Protein Lipids W_NA
What do you think about this configuration?
One more thing.
Since I have already run several simulations at higher temperature (328K), and i found that the protein, as well as the lipids 'behaves normally', should i have to run again these systems, adding in the *mdp the comm-grps line?
Thanks for your help Manuel,
Best,
Carlos
Last edit: 8 years 8 months ago by Navarro.

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4755 by mnmelo
Replied by mnmelo on topic Regarding temperature
Great that you found the cause.

Now, the most obvious solution -- which I failed to mention -- is to add the so-called 'antifreeze' water particles (molecule name 'WF' in the martini itp). These are waters with a slightly larger radius that break the packing and prevent solidification. You should replace about 10% of your waters with these.

Regarding the need for separate COM motion removal:
You say your protein adsorbs onto the headgroups. I'd say this is probably enough that you don't have to worry, unless you see the protein sliding on the surface dragged by the aqueous phase. If you're studying the system from a point where the proteins starts out away from the membrane, then you might consider separately removing COM motions at least until the protein lands on the bilayer.

Here's why:

The potential problem when the COM motion is globally removed is that the phases might start sliding relative to each other while the global COM stays the same. The respective thermostats will see this translation as an increase in temperature and proceed to scale velocities down. However, this downscaling happens on all d.o.f., not only the phase translation one. The result is that heat (velocity) will effectively build up in that translational d.o.f., while the rest of the phase's d.o.f become frozen (a flying ice cube).

This problem is averted if there is effective momentum transfer between phases. A bilayer on its own will slide relatively easily through an aqueous phase. If, however, it has an embedded protein sticking out, this protein will couple both phases' movements and the ice-cube-effect is then probably negligible.

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago #4756 by Navarro
Replied by Navarro on topic Regarding temperature
Loud and clear.
So you recommend me to set up my systems at 300K with the COM removal for the entire simulations, right?
I'll have to start reading /studying about the "flying ice cube" effect ASAP.
Best,
Carlos

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4757 by mnmelo
Replied by mnmelo on topic Regarding temperature
My main recommendation is to use the antifreeze water.

And yes, it might be good to remove COM motion separately, especially since you're interested in the temperature effect, and the consequence of the flying ice cube is a sort of misreported temperature.

Manel

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago - 8 years 8 months ago #4770 by Navarro
Replied by Navarro on topic Regarding temperature
Updating.
I was able to run several simulations at 300K (DFMG-POPC, FPMG-POPC, OPMG-POPC, DPMG-POPC, pure DPMG and OPMG without protein) but I'm stuck simulating pure DFMG (all the system are in a cubic simulating box of 10x10x10).
For some reason, during the simulations the X and Y axys start to shrink, getting different types of error messages, like:
Step 7149080: The domain decomposition grid has shifted too much in the Y-direction around cell 1 3 0. This should not have happened. Running with fewer ranks might avoid this issue.
For more information and tips for troubleshooting, please check the GROMACS

or
The X-size of the box times the triclinic skew factor (1.000000) is smaller than the number of DD cells (4) times the smallest allowed cell size (1.800000)

This is the *mdp i'm using for all systems:
Warning: Spoiler! [ Click to expand ]

I tried launching the simulations with 32, 16 and 8 cores, without luck :/
What could be the reason for this strange behaviour?
Best,
Carlos
edit- For some reason i cannot copy the whole content of my input *gro file, so here is it of you need to check it:
cl.ly/0X042k0S0A3K
Last edit: 8 years 8 months ago by Navarro.

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4771 by mnmelo
Replied by mnmelo on topic Regarding temperature
That seems a bit odd, especially since you got quite far in your simulation before it happened.

My bet is that you need a larger tau-p (or a lower compressibility). I'd try a tau-p of 5ps, or a traditional compressibility of 4.5e-5.

If this still fails and you keep consistently getting errors a bruteforce way to debug this and other problems is to output every step of a simulation (nstxout=1). The idea is to get a full frame trajectory leading up to the crash, and then trying to see what causes it. You must run it only for a couple thousand steps at a time (use -maxh 0.05, or similar), clean up, and then continue if it hasn't crashed yet (use -cpi and -noappend) otherwise you'll quickly create a LOT of useless trajectory data.

Other things I'd do, unrelated to solving the problem:
-Increase the dt to 0.02. This might even make the system more unstable, but Martini stuff should be able to run at dt=0.02ps, except for some special cases. (Unless, of course, you are simulating something special).
-Set nstcomm to, at least 100. I think grompp does it for you. The point is you don't need to burden the engine with comm removal so frequently.

Good luck,
Manel

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago #4772 by Navarro
Replied by Navarro on topic Regarding temperature
Hi Manuel,
Thanks for the quick reply :)
I'll try your 2 suggestions ASAP.
But i do have another questions. Considering i'll change the taup and/or compressibility for just one simulation, this change will no altere a future proper comparison with the other simulations?
Thanks again!
Carlos

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4773 by mnmelo
Replied by mnmelo on topic Regarding temperature
Well, yes it will. This will affect mainly membrane fluctuations. How important these are is up to you.

On the other hand, since that crash happened well into 70ns of simulation, it might just be some rare instability. If you rerun the simulation with the same parameters you might get it to finish cleanly.

Manel

Please Log in or Create an account to join the conversation.

  • Navarro
  • Navarro's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
8 years 8 months ago #4774 by Navarro
Replied by Navarro on topic Regarding temperature
Thats the problem. Since i'm measuring also the Area per lipid and membrane thickness of the lipids, and how its fluidity can alterer the adsorption of the proteins into its headgroups i must analysed all the bilayers using this same parameters.
I tried reruning the simulations several times with the same issues. The only difference, it that they crashed at different times, but non of them has passed at least the 100ns.

Please Log in or Create an account to join the conversation.

More
8 years 8 months ago #4775 by mnmelo
Replied by mnmelo on topic Regarding temperature
Then I'd say you use that system to find the conditions that work well, and repeat all others with the same ones.

If you do get the dt=0.02 to work, it's already half the simulating you'll have to do. (don't forget to double the trajectory and energy output frequencies, then. Forgot to mention that earlier).

Manel

Please Log in or Create an account to join the conversation.

Time to create page: 0.125 seconds