Re: [AMBER] GPU-accelerated thermodynamic integration available in Amber

From: Jason Swails <jason.swails.gmail.com>
Date: Wed, 13 Sep 2017 22:52:00 -0400

On Wed, Sep 13, 2017 at 4:16 PM, Daniel Mermelstein <
danmermelstein35.gmail.com> wrote:

> Hi Adam,
>
> Which mutation? A mutation with a change in net charge? I've never tried,
> but I THINK the main concern is that the pme reciprocal calculation won't
> converge, so in principle you could just have an ion that changes along
> with lambda.
>

​PME with a net charge in most programs is no problem -- it applies what's
referred to as a "net neutralizing plasma" to ensure the sum converges.
The way this is done in practice, I believe, is to simply omit the k=0 term
in the reciprocal sum (which is the divergent term in charged cells). This
is physically equivalent to taking that net charge and smearing a
neutralizing countercharge equally over the mesh.​

The effect this has on computed thermodynamic properties has been
investigated before (with one of the earlier papers being coauthored by Tom
Cheatham -- see http://aip.scitation.org/doi/abs/10.1063/1.476320 for more
information.

If I recall my conversation with Bernie Brooks (another of the coauthors on
that paper) a few years ago -- the neutralizing charge applied to the mesh
actually induces a kind of artificial attraction between the particles in
the system and the points on the mesh (after all, opposite charges
attract), which reduces both the computed pressure and free energy compared
to if this net charge and its correction did not exist in the first place,
I think.

Naturally the size of this effect depends heavily on the size of the box,
with this effect being far more muted in large unit cells. Keep in mind
that most TI calculations are performed using a thermodynamic cycle, so
assuming that care is taken when defining this cycle (e.g., the net charge
states of the model compound mutation and the mutation in the system of
interest are the same with similar solvent environments), then these
effects largely cancel when computing the desired free energy.

Anyone who has done this before or can give a more definite answer feel
> free to chime in!
>

​Just about any pKa calculation utilizing TI does this. I don't have
access to journal articles anymore to find you specific citations, but I
know I've read several in a past life.

All the best,
Jason

-- 
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 13 2017 - 20:00:03 PDT
Custom Search