Re: [AMBER] Dummy atom and PMEMD

From: Jason Swails <jason.swails.gmail.com>
Date: Sun, 5 Oct 2014 21:02:01 -0400

On Oct 5, 2014, at 4:11 PM, Miroslav Krepl <krepl.seznam.cz> wrote:

> Dear Amber users,
>
> by any chance, did any of you eventually figure out a solution to this
> problem:
>
> http://archive.ambermd.org/201208/0452.html
>
> ?
>
> I can confirm that two years later, the pmemd.MPI in amber 14 is still
> doing the same thing for my dummy atoms, i.e. reporting message
>
> "ERROR: PMEMD does not support single H residues! �
>
> for my dummy atom and then crushing down...
>
> I have been using transformation into dummy atoms in the newly
> implemented one-topology Thermodynamic integration in pmemd.
>
> ---
>
> I think that I managed to identify the cause of the problem...
>
> I tried changing different parameters of the dummy atoms using parmed.py
> and I eventually figured out that the zero atomic mass is somehow
> offending the pmemd...
>
> It seems that pmemd thinks all atoms with smaller mass than hydrogen
> (e.g. the zero-mass dummy atoms) are hydrogen!!!!
>
> I tried setting the dummy mass to "0"... simulation crushed with the
> message.

You can�t integrate atoms with a mass of 0. Consider Newton�s equation � F=ma. If the force is non-zero, the acceleration is infinite. It�s not at all unusual that an MD code doesn�t handle atoms with a mass of 0 (virtual sites are different beasts altogether).

> I tried setting the dummy mass to "1"... simulation crushed with the
> message.
>
> I tried setting the dummy mass to "2"... simulation run completely fine
> (albeit obviously generating wrong results because the dummy atom now
> has non-zero mass).

Define �wrong results�. Equilibrium properties are independent of mass. It messes up the kinetics (how much depends on how �wrong� the masses are). To my knowledge, dummy atoms have usually been given the mass of the atoms they are �replacing�. But read on below�

> ---
>
> Unless I am missing something else, I'd like to report this behavior
> (identifying atoms with lower mass than hydrogen as hydrogens) as
> possible bug of the pmemd.
>
> Right now it is limiting the applicability of the new TI implementation
> (which is pretty great otherwise).

I�m curious as to why you think that you require dummy atoms for softcore TI. They shouldn�t be necessary for either pmemd or sander (only if you do not use softcore in sander should they be necessary).

All the best,
Jason

--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Oct 05 2014 - 18:30:02 PDT
Custom Search