Re: [AMBER] free energy profile

From: Lars Skjærven <>
Date: Thu, 10 Jun 2010 00:42:25 +0200

Hi Thomas,
Great. Thanks for response. I made myself a simple reaction coordinate
to start with: the angle between three domains. Calculated that with
ptraj, and measured min and max values of this.

As input to WHAM, I used the measured angles, and 0 as spring constant
for the biasing potential:
"wham Pval 74.1 131.5 200 0.0001 300.0 0 angle_metadatafile angle.pmf"
where 74 and 131 are min and max angles, 200 is the number of bins,
0.0001 is the wham convergence tolerance, and 300 is the temperature.

Does this sound reasonable in order to get some sort of free energy
profile as a function of the domain angles?
I guessing I'm a bit off here since you say "by hand", and two-window..?

Kind regards,
Lars Skjærven

On Wed, Jun 9, 2010 at 9:16 AM, <> wrote:
> Hi Lars,
>> Given two long MD simulations (NVE) of the same protein that sample
>> two different conformational spaces (e.g. an opening vs a more fixed
>> conformation due to nucleotide binding), would it be possible to
>> calculate a free energy profile that can be related to the
>> conformations in the trajectory?
> If the two subspaces of the conformational space simulated in the two MDs
> overlap, then yes, you should be able to do something like a two-window
> umbrella sampling by hand. All you need to do is define a reaction
> coordinate and make histograms of your structures along that. Technically
> you would just be doing US with a zero additional potential.
>> Is this only possible by performing an umbrella sampling, or would it
>> make sense to extract internal energies and solvation terms and relate
>> this to the conformational changes somehow?
> The energies already did their work in my opinion, the probability to
> observe a given conformation should already be related to its free energy.
>> In regard to umbrella sampling, would it be reasonable to perform it
>> to drive large scale conformational changes? i.e. by driving an
>> opening of the protein (300? residues) based on angle/distance
>> restraints on the structure (where the restraints are set between
>> center of masses of two/three domains in the protein). Or should it be
>> limited to smaller conformational changes, as atom-atom restrains?
> I believe, as long as you can get converged simulations (a big 'if'!) any
> change can be analyzed that way. The difficulty may be in finding a good
> reaction coordinate to drive your system along. Maybe for a case like
> this, a NEB simulation would be preferable?
> Kind Regards,
> Thomas
> Dr. Thomas Steinbrecher
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Wed Jun 09 2010 - 16:00:04 PDT
Custom Search