# Re: Unusual WT-Metadynamics gaussians heights

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Tue Nov 13 2018 - 13:50:47 CST

Hi Bruno, I ran a quick test with a very low biasTemperature, and was able
to see the local Gaussian weight (last column of the hills trajectory aka
HILLS file) decrease monotonically as the hills are added near the same
point. Can you paste a snippet of your HILLS file to show how far into the
trajectory you see the inconsistency?

FYI, check that the explicit value of hillWidth is what you wanted: this
number measures the width of a Gaussian in number of grid points. In the
most recent version of the manual I shortened that entry for a quicker
read, but even in the previous version the last sentence is explicit:
The Gaussian width along each colvar, *2\sigma_{xi_i}*, is set as the
product between this number and the colvar's parameter *w_i* given by width
(see 10.2.1); such product is printed in the standard output of NAMD. ... *
Note: values smaller than 1 are not recommended. *

The default value of hillWidth usually works well, and you should set the
spatial resolution of the PMF simply through the width keyword (i.e.
identically to what you would do with ABF).

If you wish, in a future version you will be able to provide sigma and grid
spacing as independent parameters (i.e. not using a default number for
their ratio). That is the syntax followed in the metadynamics
implementations of Parrinello's group and alumni, including the most recent
ones in PLUMED 2. The current syntax in Colvars will remain supported, but
easily.

Giacomo

On Tue, Nov 13, 2018 at 11:16 AM Bruno Cuevas Zuviría <bruno.czuviria_at_upm.es>
wrote:

> Oh it's right, the file I uploaded is not the right one, I have just
> realised. Anyway, my output at initialization confirms that well-tempered
> was on.
>
> https://gist.github.com/BrunoCuevas/67dddc26c858f74701b67d230dcd0303
>
> I compiled my namd from source code pushed the 2018-07-23 in Linux64
> machine.
> El 13/11/2018 a las 14:44, Giacomo Fiorin escribió:
>
> Hi Bruno, can you post the standard output at initialization time? From
> the configuration file that you used, you do not have the wellTempered
> option on: therefore, I'm not sure how the weight could change.
>
> If you are indeed observing Gaussians with variable height given your
> configuration, then please report also the code version you are using.
>
> Giacomo
>
> On Mon, Nov 12, 2018 at 5:27 AM Bruno Cuevas Zuviría <
> bruno.czuviria_at_upm.es> wrote:
>
>> Hi all
>>
>> I am working in a ligand-protein system using well tempered
>> metadynamics. It was on my understanding that gaussians height should
>> decrease according to the underlying free-energy landscape that is drawn
>> by the gaussian deposition, so gaussians should never be larger than the
>> maximum value that was set.
>>
>> However, I am reading from the hills file values higher (~0.3kcal/molA)
>> than what I set (0.1kcal/molA). Why is this? Could it be related with
>> the fact that my colvar has run out of the grid? Or with the fact that I
>> am running a re-started colvars simulation?
>>
>> The colvars config file I am using is this one:
>>
>> https://gist.github.com/BrunoCuevas/3441584d7fc944bb1b97c2dbd90ce487
>>
>>
>> --
>> -------------------------------------------------------------------------
>> Bruno Cuevas Zuviría
>> Centro de Biotecnología y Genómica de Plantas - UPM/INIA
>>
>>
>> ---
>> El software de antivirus Avast ha analizado este correo electrónico en
>> busca de virus.
>> https://www.avast.com/antivirus
>>
>>
>
> --
> Giacomo Fiorin
> Associate Professor of Research, Temple University, Philadelphia, PA
> Contractor, National Institutes of Health, Bethesda, MD
> http://goo.gl/Q3TBQU
> https://github.com/giacomofiorin
>
> --
> -------------------------------------------------------------------------
> Bruno Cuevas Zuviría
> Centro de Biotecnología y Genómica de Plantas - UPM/INIA
>
>
>
> de virus. www.avast.com
--