Re: FEP difficulties between two ion species

From: Jérôme Hénin (jhenin_at_ifr88.cnrs-mrs.fr)
Date: Tue Apr 24 2012 - 07:19:45 CDT

Hi Jeremy,

On 23 April 2012 23:38, Jeremy Parlin <jeremyparlin_at_gmail.com> wrote:
> I have a system with an ion bound in a protein pocket, with one distance
> restraint on it set via colvars. I am trying to perform an FEP simulation,
> where that ion changes species, while the colvar restraint stays applied.
>
> I've set this system up, with both ions present in the topology file, having
> the same initial coordinates, with the old ion set to annihilate and the new
> one to exnihilate. This executes successfully, however, in the early stages
> of the simulation, the new ion swims out of the pocket, since it experiences
> no protein interactions holding it there. To remedy this, I've tried to 'tie
> together' the two ions. I've tried to do this either with a colvar distance
> constraint between the two ions (restraining the distance to zero), or by
> creating a virtual bond in the PSF file (adding a corresponding term to the
> parameter file). In either case, the simulation fails on the very first
> step, with both ions having infinite velocity.

This indicates that something is wrong in the way the simulation is
set up. The two ions should not interact with each other.

> Thoughts on how to do this successfully? I'd really like to the exnihilated
> and annihilated ions to be in effectively the same position as one another
> at all times, but both to be allowed to fluctuate together within the
> binding pocket subject to my colvar restraint.

The strategy you describe is perfectly fine and should yield the
result you want.

Now, the problem is probably with your "alchfile" - check that the
alchemical flags are properly set to -1 and +1 on the two ions, and
check NAMD's standard output at startup to see if everything is parsed
as intended. If you can't find the issue, send us a compressed log
file.

Last thing: make sure you use a recent version of NAMD, ideally the
latest, as these things are more automated in recent versions.

Jerome

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:21:54 CST