The difference between sum_hills and reweighting (using the metadynamics bias with/without the upper wall bias)

1,758 views
Skip to first unread message

杨欣

unread,
Mar 29, 2021, 12:03:38 AM3/29/21
to PLUMED users
Hi Plumed Users,

In this tutorial (https://www.plumed.org/doc-v2.5/user-doc/html/ves-lugano2017-metad.html),  the FES estimated from sum_hills is same with the reweighted FES using using the metadynamics bias (in section "Reweight the simulation"), but diefferent with the reweighted FES using both the metadynamics bias and the upper wall bias.

According to my understanding, we have to correct the free energy surface for the walls when using walls in our metadynamics simulation. Going back to this example in the tutorial, does this mean that using sum_hill will get wrong results?

Thanks in advance.

Giovanni Bussi

unread,
Mar 31, 2021, 3:55:52 PM3/31/21
to plumed...@googlegroups.com
Hi,

it depends if you want to "unbias" for the presence of the walls.

If you try to do it, you will notice that this will result in large statistical errors on the walls because you actually spent less time there. Also notice that the same is true if you run a plain MD with walls: to remove a posteriori the bias of the walls, you will actually increase the weights of the frames on the walls, that are few (because the system was discouraged to go there), making errors very large.

Typically if you used walls it is because you didn't want the system to go somewhere, so it does not make sense to compute the free energy *exactly there*.

There is a special case where you can reconstruct what happens past the walls without the system explicitly going there, e.g. when you limit the conformational space for a ligand so that it does not explore the full box. In this case, the result would be dependent on the ligand concentration (in the experiment) and thus, doing the math right, you can obtain the correct free-energy at a fixed bulk concentration of the ligand.

Yet another case is when you add walls to make sure something nasty wouldn't happen, but it would not have happened anyway. The walls will never be activated and their unbiasing will lead to no change.

Hope this clarifies things more than confusing them... overall the answer is problem dependent I think.

Giovanni


--
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/e512e21d-fcd2-4226-8750-22d71c99132cn%40googlegroups.com.

pat...@seas.upenn.edu

unread,
May 25, 2021, 12:49:47 PM5/25/21
to PLUMED users
Hi Dr. Giovanni,

Does the sum_hills action itself involve the reweighting for the free energy landscape since the unbiased free energy landscape is what we actually care about? In other words, does sum_hills give the unbiased free energy landscape? (like for e.g. in the tutorial  https://www.plumed.org/doc-v2.6/user-doc/html/belfast-6.html  , nothing is mentioned on reweighting in the plumed script and sum_hills is used to estimate the free energy landscape, is that unbiased free energy landscape?)

Is it correct to say that a biased or non-reweighted free energy landscape would be required to get the estimate of weights on the CV space in the case that we have to calculate some property other than free energy landscape as an ensemble average?

Many Thanks

-Keshav

Giovanni Bussi

unread,
May 25, 2021, 4:29:43 PM5/25/21
to plumed...@googlegroups.com
Hi,

there's no way to do a reweighting with sum_hills. This is just summing the Gaussian to reconstruct the free energy as (proportional to) the applied potential.

Regarding the second question I would say: no, it is not correct. Avoiding reweighting or not does not depend on what you compute, but on how you want to consider the biases that you possibly applied during the simulation.

Giovanni




Keshav Patil

unread,
May 26, 2021, 2:06:43 PM5/26/21
to plumed...@googlegroups.com
Dr. Giovanni,

Thanks for your explanation.

Best,
Keshav


Keshav Patil
B.Tech (Hons.) IIT Kharagpur
CBE PhD - 4th year, University of Pennsylvania
Radhakrishnan Lab


Reply all
Reply to author
Forward
0 new messages