Re: [AMBER] Dummy atom and PMEMD

From: Jason Swails <>
Date: Mon, 6 Oct 2014 07:26:38 -0400

On Oct 6, 2014, at 4:13 AM, Miroslav Krepl <> wrote:

> Hello Jason,
> yeah, ok. That is a very good point with the infinite acceleration.
> Haven't thought of that. The kinetics was my concern, but as you pointed
> out, it does not matter for the equilibrium properties. Ok.
> As for why I am doing this, that's a little tricky to explain :)
> Basically, I have a protein in explicit solvent and I am mutating some
> of its charged residues (Lys, Asp, etc.) into various different amino
> acids, most of them uncharged.
> This obviously changes the net charge of the system away from
> neutrality... so I figured that I could keep it the same by
> simultaneously transforming one of the explicit ions (either cations or
> anions) into a chargeless dummy particle.
> The approach with dummy particle is a little clumsy, but I haven't
> figured out a way to simultaneously transform one amino acid with SC
> while also decoupling an ion from the system…

You don’t _have_ to keep the net charge neutral for these alchemical mutations. PME handles a net charge just fine by utilizing a “net neutralizing plasma” by default that effectively smears the excess charge evenly over the whole charge grid in the reciprocal space.

Since most free energy calculations of this type require some kind of thermodynamic cycle (involving some sort of simplified “model” compound), you can get this effect to cancel out by including it in each leg of the cycle. In fact, transforming an ion introduces considerable complexity in terms of converging your final answer (since you rely on not only full sampling of the surroundings of your mutating residue, but also the full sampling in the solvent distributions around the “new” ion. And then of course there’s the added complexity involved with how you define your thermodynamic cycle to include the new ion.

That said, an alternative to using a dummy atom is to simply mutate a single water molecule into an ion (if you choose an anion like chloride, then you don’t have to worry that your water molecule is “bigger” than what it is morphing into). Water behaves largely like an ion with a multipole right now (due to the displaced charges of the hydrogens). If you set the hydrogen charges to zero (and move the oxygen charge off the virtual site and back onto the oxygen for TIP4P), then you effectively have an ion, since most water models are analytically rigid and the hydrogen atoms have no van der Waals terms.

Particles are prevented from collapsing into the hydrogens by the vdW repulsion of the oxygen atom. So you can imagine that if you change the vdW terms of the oxygen atom in a water molecule to match the vdW terms of an ion (and change its charge so the oxygen charge changes to the ion charge the hydrogen charges go to zero), it will to a _very_ good approximation look just like an ion.

Just some ideas. I would recommend looking for studies where people have done the kinds of mutations you are trying to do and see what they’ve done.

Good luck,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Mon Oct 06 2014 - 04:30:02 PDT
Custom Search