Re: [AMBER] polarizable FF in amber18 or 20

From: David A Case <>
Date: Sun, 7 Feb 2021 10:34:01 -0500

On Sun, Feb 07, 2021, Vaibhav Dixit wrote:

>I want to use a polarizable force field for MD simulation of a protein with
>non-standard residues and cofactors.

As you post points out, we don't support such simulations. There is still
some old code around, and you could probably cobble together a simulation of
a standard protein using a two-decade old force field. Structural results
would almost certainly be worse than if you used a non-polarizable modern
Amber force field. And creating polarizable models for non-standard
residues and cofactors would present a significant challenge.

The Amoeba force field is supported in gem.pmemd. But that is not its
primary purpose, and there is no GPU support. Plus, you have to set up the
system in Tinker as a start, and by the time you get all that done, you
might as well run the simulation itself in Tinker or OpenMM. Again, whether
your results "improve", compared to using something like ff19SB, will depend
on the system and what sorts of properties you are looking at.

Another alternative, that I use, is to use just OpenMM, skipping the Tinker
part. You can use the scripts below to make these simulations look more
like Amber runs, outputting netCDF trajectory and restart files that can be
analyzed via cpptraj or pytraj.

Others on the list might wish to chime in with ideas here. There is some
work in the Amber community to develop new polarizable force fields and
codes, and I don't know how far along these efforts are.


AMBER mailing list
Received on Sun Feb 07 2021 - 08:00:02 PST
Custom Search