On Tue, Apr 18, 2017, George Tzotzos wrote:
> I’m dealing with a homodimer and I’m trying to calculate the entropy of
> association of the two monomers. Each monomer has 125 residues. I’ve
> conducted 3 calculations for the last 40 ns of a 100 ns trajectory,
> using the same traj.nc, topologies and input file. The number of frames
> in each calculation was 10, 20 and 40 frames respectively.
>
> The results are summarised below:
>
> 1st calculation - 10 frames
> DELTA S: -62.8801 Total time taken: 29.583 min.
>
> 2nd calculation - 20 frames
> DELTA S: -89.2849 Total time taken: 30.667 min.
>
> 3rd calculation - 40 frames
> DELTA S: -136.5619 Total time taken: 30.714 min.
>
> The differences observed are due to the vibrational component of the entropy (in all 3 calculations, the translational and rotational components are almost identical).
>
> I’m aware of the shortcoming of such calculations but I have problems in
> understanding the very large variation in the vibrational component of
> entropy. Could it be that the time interval between consecutive frames
> is too short? If this is the case, how does one determine an “optimum”
> interval?
It takes a *long* time and a lot of sampling to converge quasiharmonic
estmates of entropies. Holger Gohlke and I wrote a paper on this a long
time ago, but I think it is still relevant:
%A H. Gohlke
%A D.A. Case
%T Converging Free Energy Estimates: MM-PB(GB)SA Studies on the
Protein-Protein Complex Ras-Raf
%J J. Comput. Chem.
%V 25
%P 238-250
%D 2004
Others on the list may know of more recent studies, but the principles remain
unchanged. You want to use as many snapshots as you can. But the key
"problem" is that low-frequency modes (which dominate the vibrational entropy)
converge slowly with simulation length.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 18 2017 - 05:30:05 PDT
