[gmx-users] Umbrella sampling restraining potential

[Justin Lemkul]

On 2/14/15 7:56 PM, Agnivo Gosai wrote:
> Dear Users
>
> The PMF generated by umbrella sampling of my system appears to be converged
> and the calculated free energy of binding is close to that in published
> literature.
> I believe a necessary check for the convergence is the overlap and
> smoothness of the histograms generated from sampling of the reaction
> coordinate. Earlier I discussed in the forum regarding my histogram spacing
> and smoothness.
> Dr. Lemkul had then pointed out the rocky nature of many histograms and the
> inadequate overlaps.
>
> previous post :
> http://comments.gmane.org/gmane.science.biology.gromacs.user/74659
>
> For my system there is no experimental study and I am resorting to trial
> and error. My pull force constant is 1000 kJ/mol/nm^2 whereas the only
> literature doing a MD study of my system ( with slightly different pulling
> coordinates) uses a very high force constant of 12600 kJ/mol/nm^2 and they
> have provided no explanation for this.
>
> I have increased simulation time of my windows from 1 ns to 10 ns and I am
> thinking of using a higher force constant in one half of my windows where
> the pulled group is closer to the reference group and use a lesser
> potential in the rest of the windows.
> However I am declined to use the 12600 units of force constant as it is
> very high and there is no validation. I found some literature using 1000 -
> 10,000 kJ/mol/nm^2 force constants for proteins.
>

Ask the corresponding author of the other paper why they did what they did. 
Often these values are chosen because they "just work" in the application.  I 
don't see why you'd be averse to 12,600 but OK with 10,000.  They're the same 
order of magnitude.  The point is to get proper overlap; if you can't achieve 
that with simply adding a few more windows, then you have to increase the force 
constant, which may be very high for regions of the reaction coordinate that are 
unfavorable.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================