[gmx-users] Re: another pull-code question

Justin A. Lemkul jalemkul at vt.edu
Fri Jul 2 12:35:00 CEST 2010

Thomas Schlesier wrote:
> For umbrella sampling you need to steps:
> 1) get the configurations for the later umbrella sampling (step 5 in 
> Justins tutorial). How you perform the pulling sim. for that isn't 
> important, you only need the structures.
> 2) umbrella sampling (step 6 in tutorial). Here you use the 
> configurations from step 1. I think here you can only use 
> "pull=distance" (i think).

Both "distance" and "position" work with the WHAM algorithm.  Only "direction" 
is unsupported.  It would be nice for this to be officially documented (if it 
isn't already) so that users don't find this out after collecting a huge amount 
of data that they can't analyze.  I know it's on the wiki (I put it there), but 
a statement in the manual is definitely warranted.


> Greetings
> Thomas
>> Thanks Thomas
>> That has made things a lot clearer, but just to be sure. If I pull the
>> molecules using pull = constraints, do I still have to run umbrella
>> sampling on the generated configurations in order to calculate the PMF?
>> Also if I use pull_geometry = direction for the pulling can I use it for
>> the umbrella sampling with the intention of then using g_wham for the
>> analysis?
>> Many Thanks
>> Gavin
>> Thomas Schlesier wrote:
>>>> Gavin Melaugh wrote:
>>>>> Hi All
>>>>> I am using COM pulling to pull two molecules together in order to
>>>>> calculate the potential of mean force and the pulling is fine. I am
>>>>> using pull = umbrella for the pulling as I am following
>>>>>  the tutorial on the website. If I use pull = constraint does this
>>>>> still
>>>>> pull the molecules together, and if so do you still then need to
>>>>> perform umbrella sampling on the configurations generated? I am a
>>>>> little
>>>>> confused  as I don't understand how the molecules pull together if 
>>>>> they
>>>>> are constrained.
>>>> For plain SMD, you can use any of the pull methods.  In the tutorial
>>>> I wrote, I explain why I used "pull = umbrella" during SMD.  Umbrella
>>>> sampling is a separate procedure, which can be done after SMD.  If
>>>> you want a free energy along the reaction coordinate, there are
>>>> several methods for doing this, among which is umbrella sampling.
>>>> -Justin
>>>>> Many Thanks
>>>>> Gavin
>>> Some additions to Justins comment:
>>> If you want to perform the umbrella sampling (or calculating the pmf
>>> form integrating the constraint forces) it doesn't really matter how
>>> you obtain the structures for the different windows (but you will need
>>> them).
>>> If you only want to perform the calculating of the pmf,
>>> "pull=constraint" it's a little easier to get these structures. You
>>> can perform your pulling so that each frame of the trajectory
>>> correspondents to one structure and then extract them from the .trr /
>>> .xtc. Also the pulling velocity doesn't really matter. If you pull
>>> with 1nm/ns and save the coordinates every ns, then the distance of
>>> your molecules will increase/decrease with 1nm between the different
>>> frames.
>>> If you use the same setup with "pull=umbrella" the change of the
>>> distance won't be 1nm but will depend on "pull_k1" the spring constant
>>> with which you're pulling. So you must "search" the structures with
>>> the right distances (for that is Justin's Perl script).
>>> The nice thing with "pull=umbrella" is that you can see qualitativly
>>> what happens during the pulling. If there is somewhere a barrier in
>>> the pmf you will see it (more or less) that the distance between the
>>> two molecules will change slower then if the is no barrier. With lower
>>> velocity and spring constant (pull_k1) you will see that better.
>>> But with "pull=constraint" you won't see that. Think the best way to
>>> understand the difference between "pull=umbrella" and "=constriant" is
>>> that the later is the first in the limit of an infinte "pull_k1".
>>> Hope that helps. Good thing could be, take a small test system and
>>> play around with the parameters and look what happens.
>>> Greetings
>>> Thomas
>> ------------------------------


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list