[gmx-users] Umbrella Sampling and histograms smoothness

[Justin Lemkul]

On 2/9/15 4:11 PM, Agnivo Gosai wrote:
> Dear Users
>
> My umbrella sampling simulations are of 1 ns time whereas Dr. Lemkul's
> tutorial advocates a 10 ns simulation time. I found that after running WHAM
> , my histograms are coarser ( rough) compared to Dr. Lemkul's. A few of my
> histograms have two peaks!!
> So should I increase the simulation time from 1 ns to 10 ns ?? I do a 100
> ps NPT equilibration and then a 1 ns umbrella sampling in each of my
> windows. The generated PMF curve is also not smooth and is discontinuous at
> one position which I think I can fix by adding more windows around that
> co-ordinate.
>
> A picture is linked : ​
>   HISTO
> <https://docs.google.com/file/d/0B-U8uULVZjfRQXcyUkI2bXhBcVE/edit?usp=drive_web>
> ​
> Kindly help.
>

You need to assess convergence like you would for any type of system.  Most of 
your peaks look rocky, so that generally indicates insufficient sampling, poor 
window setup, suboptimal restraint force constant, or some combination of those.

Don't think that 10 ns is some magic number, either.  It worked for that system 
after I did a bunch of checking and the result was converged.  Some umbrella 
sampling takes way more time.

-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

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