Transition tempered metad

[Manuel David Peris]

Hello all,

I am trying to set up a transition-tempered metadynamics.

However, I get the next error:

"ERROR in input to action METAD with label metad : keyword TRANSITIONWELL could

not be read correctly"

Could anyone help me with this?

Thank you

Manuel

My plumed input file looks like...

metad: ...
METAD
ARG=zn69
TTBIASFACTOR=6
HEIGHT=0.5
SIGMA=0.02
PACE=500
FILE=HILLS
TEMP=298
TRANSITIONWELL=10
...

[Michele Invernizzi]

Hi Manuel,

with the TRANSITIONWELL keyword you should provide the position of all the metastable basins of the system (at least two). See here for an example: https://www.plumed-nest.org/eggs/19/000/data/alanine/plumed-metad.dat.html

Also, in case you are using ttMetaD because you are looking for better convergence, and if I am allowed to do some shameless self-promotion ;-) you might be interested in this https://parrinello.ethz.ch/research/opes.html

You can find various plumed input examples on the plumed-nest, just search for OPES.

Best,

Michele

--
You received this message because you are subscribed to the Google Groups "PLUMED users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/c4949e97-597a-4e3d-b7a4-cc297fe24492n%40googlegroups.com.

[Manuel David Peris]

Hi Michele,

Thanks a lot for your answer. Yes, I wanted to try TTmetad just and see if the convergence improves against wt metad, which I run with a single walker.

I heard about your OPES advancements, it just look great. I will try OPES but first I need to recompile my plumed to incorporate the module.

best wishes

Manuel

[Manuel David Peris]

Hi Michele,

I set up the TRANSITIONWELL keyword and run ttMetaD. However I do not understand how the bias is deposited. Up to 300 ns there are multiple recrossings but  after 300 ns the CV get stuck in one minima for 300 ns. Please see the evolution of the CV-s with the time. Could you please help me to understand it?

thank you

Manuel

[Michele Invernizzi]

Hi Manuel,

It's hard to say what the issue is only from these two figures...

One possible explanation is that after the first few transitions the ttMetaD greatly reduced the hills heights, and the bias is almost constant in time, but the system found a new metastable state (maybe a slightly defected state) and remained stuck there. If this is the case, the best solution would be to change the set of CVs, in order to better discriminate also this other state. By using non-tempered metadynamics you would not remain stuck and would have many transitions, but anyway you will not be able to reliably converge to the correct free energy in a reasonable time. For reference see Fig. 3 in https://doi.org/10.1016/j.revip.2017.05.001

The fact that the bias reaches more rapidly a quasi-static regime helps a lot in diagnosing this kind of CV problems, but unfortunately does not solve it. If it is not possible to improve on the CVs, you might want to simply run longer simulations and/or use many multiple walkers. Once the bias is quasi-static you will be able to reweight and get reliable results. See also https://www.plumed.org/doc-master/user-doc/html/opes-metad.html

You should check if you actually are in the situation I described or not (are the hills heights extremely small? by looking at some other order parameter or by carefully looking at the full atoms, is something slightly different in this final state? If you rerun the simulations with different input parameters do you always see a similar behaviour?)

It might instead be a problem related to ttMetaD itself, about which I have not much experience.

Hope this helps,

Michele

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/73280b2f-597d-49e4-a860-27e5f37044a9n%40googlegroups.com.

[Manuel David Peris]

Thanks a lot Michele for your answer.

I checked and I see the hills heights really small (attached pic), so maybe it would be better to change the CV-s or try opes rather than add walkers..However, when I ran wt metad, and not ttmetad with the same CVs I dont see the system reaching that state...

thank you

Manuel

[Michele Invernizzi]

Hi Manuel,

If you freeze the bias you obtain from wtMetaD (e.g. by restarting the simulation and setting PACE to a very big number), does the simulation get stuck as with ttMetaD or does it go back and forth?

If the issue is the one I tried to describe in my previous email, then you will not be able to find a static bias potential that is a function of your CVs and that will not remain stuck. A non-tempered MetaD or a wtMetaD with a relatively low bias factor would give you transitions, but only because the bias greatly changes shape.

I have experienced this kind of problem with various systems, e.g. in this paper on ligand binding the only way to obtain efficient sampling was to add a CV describing water, and this allowed to discriminate "defective" states that are identical but for the position of one or two water molecules; or in this paper on crystallization we had to change our CVs to be able to distinguish the two structures in Fig. 1 of the supplemental material, that are identical but for a misalignment with the box, which was producing an extra "fake" (finite size effect) metastable state. In our experience, using a method like OPES that quickly reaches a quasi-static bias was helpful in diagnosing the problem and guiding the design of efficient CVs. I won't hide to you that with other systems we were less lucky and could not devise proper CVs to be used with a quasi-static bias. (It is also true that sometimes you do not need this type of accuracy, and just being able to explore the phase space already gives good scientific results, e.g. here, then do not use OPES nor ttMetaD)

Again, you should carefully check if this is actually also your problem, because there can be various other reasons for a simulation to fail!

Best,

Michele

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/3f3be211-93f8-41b5-b4f0-643b1bc035ddn%40googlegroups.com.

[Manuel David Peris]

Hi  Michele,

Thanks a lot for your answer:). 

I will check as you suggested.

best wishes

Manuel