Re: AMBER: soft-core potential in amber?

From: David Mobley <>
Date: Tue, 7 Feb 2006 06:54:33 -0800


At present, soft core potentials are not implemented in AMBER,
although, in my opinion, they are essential to get accurate free
energy differences anytime the number of vdW sites changes. (To give
one example, I was running some binding free energy calculations which
involved turning off a ligand's vdW interactions in the binding site
of a protein. The ligand was small, phenol or catechol. I could run
500 ps at each lambda value and the results would appear reasonable
(computed uncertainties reasonably small) but if I reran the
simulations, computed free energy differences could be wildly
different (as much as 4 kcal/mol, comparable to the total binding free
energy in this case). The dv/dlambda time series is plagued by really
severe shot noise, as one might expect without soft core potentials.

I don't think the problem is as bad for calculating free energy
differences where correlation times are short. I calculated the free
energies of turning off vdW interactions for my ligands in water
using the AMBER force field in AMBER (without soft core) and in
GROMACS (with soft core) and got comparable results over 1-5ns
simulations at each lambda value. But I get huge differences for the
component where the ligand is in the protein, as I described above. I
think the long correlation times of proteins exacerbate the problems
which arise without soft core.

Also, a by-product of all this is that not using soft core
necessitates use of smaller timesteps to keep the simulation from
blowing up, since the forces are really large. This obviously slows
down these calculations quite a bit.

Anyway, to reiterate: My current opinion is that it isn't safe to do
calculations where vdW sites are inserted or deleted without using
soft core potentials. I'm currently using GROMACS to do these sorts of
calculations, although I think CHARMM may also be able to do so. Since
this is the AMBER list, of course, others may disagree with me. :) My
hope is that at some point soft core will be implemented in AMBER.

Regarding your specific problem, though... Are you thinking of
computing the free energy difference of making a mutation in the prion
protein, in the hopes of getting at questions of stability?

Best wishes,
David Mobley

On 2/6/06, Lydia <> wrote:
> Dear all,
> We are interested in using the soft-core potential to calculate the
> free energy difference between wild type and mutant prion protein. I have
> just installed amber 8 and we may also have access to amber 7 at our
> collaborator's cluster but we don't know how and in which version amber we
> can use soft-core potential.
> Would anyone please tell me how I can obtain and use the soft-core
> potential in amber? Any suggestion or information is welcome. Thank you for
> your help in advance.
> Sincerely
> Lydia
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Feb 08 2006 - 06:10:06 PST
Custom Search