Re: [AMBER] 10-12 Potentials, HBOND

From: Jason Swails <>
Date: Sat, 14 Mar 2015 19:12:10 -0400

> On Mar 13, 2015, at 11:37 PM, David Poole <> wrote:
> Hello everybody,
> I've been working with a system where it would be exceedingly useful to put
> a 10-12 potential between iron and any solvent oxygen to better simulate
> water's behavior in the cavity. The parameters I am using for heme produce
> excellent geometries and I'd like to leave them intact.
> This leads me to ask you all:
> 1 - Do 10-12 potentials, via HBOND, still work in Amber14,

They should be. Your question actually triggered me to test (and fix) the 10-12 functionality in the version of Amber that’s due to be released in a month or so.

You need to compile Amber 14 with -DHAS_10_12 (which you can do by setting the environment variable AMBERBUILDFLAGS to “-DHAS_10_12” in order to get access to the 10-12 functionality.

> and are they
> supported in the CUDA implementation?

No. However, ParmEd ( <>) interfaces with OpenMM in order to run simulations on the GPU. The latest version on Github supports topology files with 10-12 terms, so if you want to run GPU simulations with the 10-12 potential, you’ll need to use ParmEd with OpenMM (note, OpenMM’s application layer itself does not support 10-12 topology files, so you need to use ParmEd). If you check the ParmEd documentation in the Amber 14 manual, you will see instructions for how to run a simulation using Amber input files with the OpenMM action.

> 2 - The syntax from the frcmod file specification simply says atom types,
> does this indicate that the 10-12 potential can be defined for any atom
> type pairs?

Yes, *but* a pair of atoms cannot have *both* 10-12 and 6-12 parameters.

> 3 - Does anyone know the function for this potential? I read that it had
> coefficients of 5 and 6 on the 10th and 12th power terms respectively, but
> I just wanted to confirm.

I’m not sure where you are getting these coefficients. But the functional form is:

A/r^12 - B/r^10

where A and B are defined between particular atom pairs. These values are taken exactly from the frcmod or parm.dat files (which list H-bond interactions between particular pairs.

> 3 - Is there some major consideration that I am overlooking that will ruin
> my results or otherwise break pmemd?

The only thing is if you expected both Lennard-Jones *and* 10-12 terms between the atoms. That is not supported by Amber.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Sat Mar 14 2015 - 16:30:02 PDT
Custom Search