Re: [AMBER] Quasi-harmonic Approximation Using a Trajectory Stripped of Hydrogens

From: Josh Berryman <the.real.josh.berryman.gmail.com>
Date: Tue, 28 Apr 2015 12:51:46 +0200

>>I was thinking if maybe I could reduce the number of frames needed for
the quasi-harmonic
>>calculation by removing all hydrogen atoms from the trajectory.

Absolutely, entropy from things like methyl-group rotation is probably
unimportant and not very well captured by this method anyway. It is not
unheard-of to throw away everything except the C-alphas, especially if
memory for the diagonalisation becomes a problem.

Also, try doing a convergence analyisis of what happens to the entropy as
you run the calculation on increasingly longer trajectory segments, it
should follow a power law which you can extrapolate to t=infinity. This
rather obscure maths paper gives an explanation of why it works, plus links
back to the simulation papers where the approach was discovered:

  Title = {Statistical inference for functions of the
covariance matrix in the stationary Gaussian time-orthogonal principal
components model},
  Author = {Dryden, IanL. and Kume, Alfred and Le,
Huiling and Wood, AndrewT.A.},
  Journal = {Annals of the Institute of Statistical
Mathematics},
  Year = {2010},



>>I started an MMPBSA calculation using these three trajectories and using
>>all 250 000 frames and at the rate at which the frames are being processed
>>right now, I anticipate at least 3-4 days will be spent just on the GB
>>calculations

The number of nanoseconds might be more pertinent than the number of frames
here. Collecting a PE snap every 2 femtoseconds would certainly be a bit
excessive, as the decorrelation time of your energies is rather longer than
this.

>>restarting MMPBSA...

sorry, never tried

Josh



On 27 April 2015 at 05:35, Laura Tociu <ltociu.princeton.edu> wrote:

> Dear Amber users,
>
> I am attempting to calculate the entropic contribution to the binding
> energy of a very large system and previously I had thought about truncating
> the system and using NMode but eventually decided against it because the
> protein has a very obvious correlated motion of its monomers and upon
> binding that motion is almost lost so the quasi-harmonic approximation
> would probably be pertinent to this case.
>
> The number of atoms being 6000 + I need at least 18 000 frames for the
> quasi-harmonic approximation. So I now have three separate trajectories
> (one of the protein, one of the complex and one of the ligand -- I wanted
> to use separate trajectories precisely because upon binding the motion of
> the protein changes), each having 250 000 frames.
>
> I started an MMPBSA calculation using these three trajectories and using
> all 250 000 frames and at the rate at which the frames are being processed
> right now, I anticipate at least 3-4 days will be spent just on the GB
> calculations, and I can't estimate how much the entropy calculation will
> take and whether everything will run smoothly or not from now on.
>
> Since obtaining this entropy is a little time-sensitive now, I was thinking
> if maybe I could reduce the number of frames needed for the quasi-harmonic
> calculation by removing all hydrogen atoms from the trajectory. This would
> reduce the number of frames needed by roughly a half. I was thinking about
> doing two MMPBSA calculations: one using a trajectory with all of the
> hydrogens present, from which I would derive the intramolecular energy and
> the solvation energy and one using a trajectory that is free of hydrogens,
> from which I would simply derive the entropy. I recall reading about
> something similar in an article but am not sure I remember correctly. Is
> this a valid procedure?
>
> If not could someone tell me if an MMPBSA calculation can be restarted if
> it dies for whatever reason and how? I know there is an option to use the
> existing input files, but I am not sure if this means that the script will
> pick up where it left off if it died in the middle of frame processing for
> example.
>
> Thank you!
>
> Laura
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 28 2015 - 04:00:02 PDT
Custom Search