On Thu, May 21, 2020, Carl Heroneme wrote:
>
> I am trying to calculate free energy of solvation using a POL3 polarizable
> water model with a polarizable solute. I am wondering if I can do this in
> Amber simply with thermodynamic integration or BAR like with a
> non-polarizable force field or if I will have to use a method that does not
> involve any alchemical transformations or perhaps a different software that
> does have this capability.
>
> I have heard that this is not possible because the TI code does not take
> into account the polarization energy into the potential for the dV/dL terms,
> but I wanted to seek a 2^nd opinion before giving up on the idea.
I don't see any theoretical reason why TI should not work with
polarizable potentials. The comment about "not taking into account the
polarization energy for the dV/dL terms" seems wrong to me: dV/dL for
linear mixing is just V1-V0, the difference in total potential energy
(including any polarization components) between the beginning and ending
Hamiltonians.
There are some practical problems one would face, however. They
arise from the fact that very little effort has been made by Amber
developers in the past 15 years on polarizable force fields. (There has
been effort on the force field side, along with some code developments;
but those efforts have not reached a stage where they have been
incorporated into releases.)
The upshot of this is that experience with polarizable force fields is
quite limited, as is testing. So, the parameters needed to run anything
like an efficient MD simulation with polarization would need to be
carefully re-examined. (The gem.pmemd code handles polarization well,
but is not set up for TI.) You'd have to experiment with the
combination of soft-core and polarization: soft-core is a "natural"
technique to use for hydration free energies, but as far as I know has
not be tested with polarizable potentials.
There are people who follow this list (Ray, Junmei) who *have* been
working on polarizable force fields, and may be able to provide some
detailed advice here. Maybe others can chime in as well -- I'm
commenting almost entirely from second-hand knowledge.
I know that CHARMM, openMM and Tinker are set up to do calculations like
this, (but probably not with POL3 without some setup effort.) I'm not
familiar enough with gromacs or NAMD to comment on their capabilities.
...good luck....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat May 23 2020 - 10:00:02 PDT
