Re: [AMBER] elevated temperature MD

From: Nicholas Musolino <musolino.MIT.EDU>
Date: Mon, 19 Apr 2010 13:31:05 -0400

> I apologize for this being a general question and not a proper "AMBER" one, but would sincerely appreciate any insight or short response/direction from the experts. I searched the literature, but I could not find a review that stated the consensus on these issues.
> How well are force fields adapted to higher temperatures? Are not the force fields generated and tested using temperatures at ~300K? I have looked in the literature and found that the helix structure appears to be over stabilized at higher temps, and so, is it safe to say that at higher temperatures (~400K and higher) protein simulations are more qualitative and exploratory than accurate in their residue contacts?

Hi Steve,

In my mind, there are two issues to consider in high temperature simulations. One is the sampling that takes place at higher temperature, and the other is force field accuracy at higher temperature.

At higher temperature, a system will tend to explore phase space more broadly, since energy differences are "scaled" by beta = 1/kT; the relative probability of visiting a particular state is proportional to exp( -U/kT) in the canonical ensemble. In other words, at higher temperature, you sample more broadly, but the tradeoff is that you have less "resolution" in phase space, which is sort of "flattened out," at least to the sampling routine.

As for force field accuracy, some force fields are parameterized by matching QM data, which deals solely with the energy from the configurational aspect of phase space, and shouldn't be influenced by temperature, as Carlos Simmerling just noted.

Other force fields are parameterized using physical properties at a specific temperature (density, diffusivity, RDF); some water potentials were created this way. In that case, I think it's harder to say... to the extent that the parameterization process might "build in" some entropic effects into the parameters, the model might not hold up at different temperatures.

If you're actually interested in high-temperature properties (as opposed to being "really" interested in properties at 298K, and turning to high temperature for increased sampling and decreased simulation time), I suppose you could look for papers on high-temperature protein simulations, to see whether properties are accurate. Taq polymerase comes to mind.

NOTE: Carlos Simmerling replied as I was completing this e-mail, making similar points as the above; sorry if this is duplicative.

Best regards,
Nicholas Musolino

Nicholas Musolino
Ph.D. candidate, Department of Chemical Engineering, MIT

AMBER mailing list
Received on Mon Apr 19 2010 - 11:00:02 PDT
Custom Search