I've always had trouble getting reproducible free energies for
transformations which involve disappearing atoms, even with klambda>4.
The problems tend to get worse when the number of atoms being
disappeared increases. Maybe this is the same problem you're seeing.
In my case, I found results were sensitive even to the starting seed,
not just the starting structure.
Ultimately, I decided the solution to the problem was soft core
potentials (as described by Beutler et al in the early 1990s) but
these aren't implemented in AMBER.
David Mobley
UCSF
On 5/11/06, Pan, Yongmei <ypan2.email.uky.edu> wrote:
> Hi, David and guys:
>
> I use gibbs, amber7 to calculate free energy change from PHE to ALA in a 500 residue protein.For the ten different trajectories extracted from a stable MD simulation (rms difference <1.5), the perturbation energy varies from 4kcal/mol to 8kcal/mol. So why the perturbation from PHE to ALA is so unstable, and it seem that the result depends heavily on the starting structure. this case contrasts with those of SER to ALA, or ALA to GLY perturbation, which are much more stable among trajectories.
>
> The second question is, why the perturbation for SER to ALA always fail, due to the shake failure? I also tried "electric decoupling", but the result was not good.
>
> The way my perturbation is: The direction is from big group to small one, with diminishing atoms replaced by dummy atom. All the bond and angel parameters involving dummy atoms keep the same during the perturbation, while the mass, charge and non-bond parameters are set to zero for the dummy atoms.
>
> I appreciate if anybody can give me advice.
>
> Yongmei Pan
> College of Pharmacy, University of Kentucky
>
>
>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun May 14 2006 - 06:07:14 PDT
