Re: [AMBER] Turning off non-bonded interactions for specific atomtype pairs

From: Jason Swails <>
Date: Tue, 11 Oct 2011 17:46:00 -0400

On Tue, Oct 11, 2011 at 4:49 AM, Marc van der Kamp <
> wrote:

> PS
> I just thought of another way of doing this, which wouldn't require
> changing
> the code:
> I could introduce "bonds" (with zero force constants) between all the
> hypothetical particles, so they would be in the exclusion list.
> If there are a lot of particles, however, this would lead to a whole load
> of
> bonds (i.e. between all possible particle-particle pairs), and this may
> affect performance quite drastically...
> So it would probably still be better to modify the exclusion list by
> changing the code.

The performance difference is not really going to be pronounced, IMO. Take
a look at the bonded force calculation time -- it is next to nothing
compared to the non-bonded interactions. The biggest problem here is that
the exclusion list is not *only* the list of bonded partners. It is also
the list of angle and dihedral partners. When you define additional bonds,
you also (perhaps inadvertently) define additional angles and dihedrals
which would make the exclusion list expand far more than you would expect.

You can mediate the effect of added dihedrals by defining their new scaling
factors to be 1.0, but that's an added level of complexity.

I'll respond to your first email shortly with regards to hacking the
topology file.


> --Marc
> On 11 October 2011 09:42, Marc van der Kamp <
> >wrote:
> > Hi all,
> >
> > For a future project, I am thinking of using hypothetical particles (with
> a
> > VdW radius and partial charge) that interact with a solute, but NOT (or
> > perhaps very little) with each other. The particles would have their own
> > atom types.
> > I don't think this would be possible without changing the code (i.e. by
> > hacking the prmtop or so), but please chime in if this is not the case.
> >
> > Probably the cleanest way to change the code is to modify the non-bond
> > exclusion list (or perhaps introduce an inclusion list).
> > I guess this is quite similar to what happens in QM/MM, where all the MM
> > VdW and electrostatic interactions between the QM atoms are turned off.
> (So
> > perhaps an alternative would be to introduce an MM force field as a "QM
> > method" and put the particles in the qmmask, but I think this may be
> quite
> > tricky to change in the code...)
> >
> > To start, I probably won't try using PME as this may make things more
> > complicated. I also don't care too much about affecting performance at
> this
> > stage.
> > As I am pretty much a novice when it comes to the AMBER code, it would be
> > great if someone could point me to where I could think of making changes.
> > Also, any ideas about different ways of doing this will be highly
> > appreciated.
> >
> > BTW, I did find an old thread in the archive that has some relevance
> > (although not quite the same, as I do want full non-bonded interactions
> > between the particles and the solute), but that may well be outdated as
> it
> > is from 2004:
> >
> >
> > Many thanks,
> > Marc
> >
> >
> _______________________________________________
> AMBER mailing list

Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER mailing list
Received on Tue Oct 11 2011 - 15:00:02 PDT
Custom Search