From: Aron Broom (broomsday_at_gmail.com)
Date: Fri May 11 2012 - 16:38:03 CDT
You have not done anything wrong. The PMF SHOULD keep decreasing and NOT
level off, because if you had a box of water as large as you have, it would
be more likely to not be bound, due to entropy (you have a very low
concentration of methane). Running ABF, the default is to apply a jacobian
correction to the distance colvar, which is generally what one should do.
So, the point being, your non-flat PMF is correct for your system, but you
want correct for the fact that you are not interested in the PMF at the
concentration of methane in your giant box, but rather, for the standard
state of 1M ligand. So, you just need to apply the same jacobian
correction that ABF is applying.
To do this, you will simply add: 2*Kb*T*Ln(r) to the value of the PMF at
each distance, where r is the distance, Kb is the boltzmann constant in
kcal/mol (0.001987) and T is your temp in kelvin. After adding this to
current PMF value, your long distance portion should level off.
On Fri, May 11, 2012 at 5:08 PM, Shaon Chakrabarti <shaonc_at_gmail.com> wrote:
> Dear NAMD experts,
> I am running into a problem while computing the Methane-Methane PMF
> using NAMD (Umbrella Sampling and then WHAM). After the contact minimum
> and solvent separated minimum, the PMF should level off, but the PMF I'm
> just keeps decreasing instead of becoming almost constant! I have just not
> been able to
> figure out the problem, and would be really grateful for any ideas.
> I am using an all atom representation of methane, and TIP3 water. I am
> using a
> 2fs time-step and have a 12 Angstrom cutoff for non-bonded interactions.I
> have periodic
> boundary conditions,PME and wrapAll switched on. For the harmonic biases,
> I am using
> the Colvars module in NAMD. I minimize for 20000 steps, then equilibrate
> the system
> (without any bias) for 1ns, after which I use the various biases to
> generate sampling. All
> of this is at 300 K temperature.
> 1) I have tested my WHAM code with other codes available on the web
> Alan Grossfield's code). All of them give the same wrong result. Hence
> I am almost
> certain that the mistake is not in the WHAM portion, but in generating
> trajectories using NAMD.
> 2) I have tried using water boxes of side varying from 32 angstroms to 44
> angstroms. Every
> time I get the same problem.
> 3) I used a variety of spring constants--1,2,6,8 kcal/(mol.A^2) to get
> varying degrees of overlap
> of the histograms. I have also tried placing the harmonic centers at
> different separations, but
> to no avail.
> 4) I have tried using 5000 steps (10000 fs) for decorrelation, and 40 ns
> trajectories for sampling.
> Increasing decorrelation time and sampling time doesn't change the
> 5) I have used ABF to recompute the PMF--that does give CORRECT results.
> So I guess I'm not
> committing any stupid error while building the methane-methane+water
> At this stage I'm really at a loss about the source of this problem. It
> would be of immense
> help if someone could point out where I'm making a mistake.
> Thank You,
-- Aron Broom M.Sc PhD Student Department of Chemistry University of Waterloo
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:31 CST