Re: Alchemical Free Energy Perturbation

From: Jérôme Hénin (
Date: Tue Feb 28 2006 - 09:03:55 CST


In your case, the dual-topology approach may not be the best-suited, because
the partly environment-decoupled ions are free to wander about the simulation
box and clash with the solvent.

What you could do is emulate a "single-topology" perturbation by keeping both
your ions fixed (NAMD fixedAtoms), in the very same position. (Disclaimer:
although I can see no reason why it should fail, I have never tried that sort
of trick).

Please let us know how it works.


Le Samedi 25 Février 2006 07:40, Jawahar Neelankatan a écrit :
> Hello,
> I had a few questions about alchemical free energy perturbation
> simulations. I am getting strange and inconsistent results for an
> alchemical free energy perturbation simulation that I'm doing. I am
> transforming a K+ ion to an Na+ ion in a box of ~200 methanol molecules
> (solvent). So basically the only atoms involved in the transformation in
> the FEP simulation are K (initial state, vanishing) and Na (final state,
> appearing) and thus they are prevented from interacting with each other by
> a modification (exclusion list) I made in the !NNB (non-bonding) section of
> the psf file, using the alchemify program. The K and Na ions are however
> allowed to interact with the bulk solvent (methanol) molecules. I also have
> a .fep file defined that contains the ~200 methanol molecules, K ion and Na
> ion. I wasn't sure whether to give both K and Na ions the exact same
> coordinates or not. I didn't think it'd make a difference since they aren't
> interacting to begin with. I went ahead and gave both of them the exact
> same coordinates in the .fep file. I then ran a 10000 step (2 fs per step)
> simulation with 5000 steps for equilibration and 5000 steps for sampling. I
> made delta_lambda = 0.025 to allow 40 windows for the growth of lambda from
> 0 to 1. The results I get don't bear much similarity to experimental
> results and in the .fepout output file, I'm seeing a few unrealistically
> large values (999999999) for the vdW(l+dl) i.e. the van der Waals energy
> for the lambda2 state in each window. Is this normal ? I modified the .fep
> file so that K and Na don't have the same coordinates (as they did in the
> first simulation) but then I get the 'constraint failure in RATTLE
> algorithm' error. However, I inserted a 'minimize 1000' command to minimize
> the system before equilibrating and subsequently sampling and that error
> doesn't occur anymore but I still get a few unrealistically large values
> for vdW(l+dl) and erroneous values for dG.
> The configuration file and a small section of the .fep file are attached to
> this email. Does it matter that I'm using the OPLS force field ? Any
> suggestions are welcomed. Thanks.
> -Jawahar
> ---------------------------------
> Yahoo! Mail
> Use Photomail to share photos without annoying attachments.

Jérôme Hénin
Equipe de Dynamique des Assemblages Membranaires
Université Henri Poincaré / CNRS
Tel : (33) 3 83 68 43 91        Fax : (33) 3 83 68 43 87

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:43:20 CST