Re: [AMBER] Getting free energy (difference) from GPU simulations

From: Jason Swails <>
Date: Thu, 4 Aug 2016 14:09:44 -0400

On Wed, Aug 3, 2016 at 7:18 PM, yunshi11 . <> wrote:

> Hello List,
> >From the AMBER 16 manual, I understand that thermodynamic integration and
> free energy perturbation are not supported for GPU simulations. Does this
> mean I can only calculate free energy (difference) from steered MD and
> umbrella sampling when using GPU?
> There seems to be no mentioning of linear interaction energy or
> temperature-based REMD?

‚ÄčT-REMD, steered MD, umbrella sampling, and TI are all useful for different
things. And T-REMD is really for enhancing your sampling, not for doing
free energy calculations in the sense that TI does. Steered MD and
umbrella sampling give you potentials of mean force, so unlike TI you need
to have a physical "reaction coordinate" picked out ahead of time. LIE is
really a post-processing technique applied to a simulation you run.

The closest to TI that you can currently get with pmemd.cuda is H-REMD
(Hamiltonian-based REMD), but you have to prepare the Hamiltonian
modifications yourself (using, for example, parmed to make the
modifications you want). This in effect calculates the free energy via the
Free Energy Perturbation method.


Jason M. Swails
AMBER mailing list
Received on Thu Aug 04 2016 - 11:30:04 PDT
Custom Search