[gmx-users] free energy questions

[David Mobley]

Murat,

> 1) Trivial questions first: It is advised to use PME with opls, is it possible
> to use PME in vacuo? As far as I know PME depends on periodicity, which is
> shut-off in vacuo.

Generally not a good idea to have a periodic system in vacuo, unless
you are sure that this is what you want, since there will be long
range interactions between the system and copies of itself that are
unshielded by solvent.
>
> 2) Should I extract dv/dl values from my .edr file to get dG? Can I use dgdl.xvg
> file? It is trivial but I am confused about this procedure.

Either one.

> 3) I have smooth simulations when init_lambda < 1. But When I get to lambda =1,
> I sometimes get enormously high energy values (in a fluctuating manner) for my
> system. What could be the problem?

If you are (a) not using soft core (sc_alpha=0), and (b) changing LJ
interactions, this is the reason. Indeed, dv/dlambda is actually
singular at lambda=1 if you are disappearing particles. (See the
Beutler and van Gunsteren soft core paper from the early 90s which has
been mentioned before on this list).

> 4) As you see in the mdp file, I used sc_alpha = 0. Otherwise, my system gives
> out error. Is it bad to use sc_alpha =0?

Only if you are disappearing/inserting particles. If you're getting an
error, it probably means you have something else wrong.

> 5) Should I use FE parameters in my energy minimization simulations? Normally, I
> used my minimized lambda=0 structure as input for other lambda simulations
> without doing further EM. I also tried to include lambda values in my EM files
> and I got junk results (.gro files and energy values).

Usually it is a good idea to minimize at each lambda value, at least
for stability purposes.

David