You are here

How the Rosetta score (dG) is actually related to dH (enthalpy) and dS (entropy)?

5 posts / 0 new
Last post
How the Rosetta score (dG) is actually related to dH (enthalpy) and dS (entropy)?

Dear Friends,
As we know, dG=dH-TdS. In Rosetta total score, the dG is broken down into a range of contributions with different weights. Can we assign some of those contributions to dH and some to dS? How does the temperature T play the role in Rosetta total score? Thank you very much.

Yours sincerely

Post Situation: 
Thu, 2014-12-04 13:16

The short answer is: none of us know.

There is a "temperature" involved in Rosetta's Monte Carlo simulations (in the kBT factor). The only "temperature" present in the score of a static structure is what temperatures might have been used in any simulations used to decide what weights to put on that scorefunction. That sentence is vague because the connection is vague.

The crystal structures used as references to build the scorefunction are mostly cryo xray (for higher resolution), so in some sense Rosetta thinks it is at very low temperature.

A single structure has no concept of entropy, and thus no delta S. We can address entropy with ensembles, and there are some scorefunction terms (not used universally) that attempt to explicitly or implicitly consider entropy.

Some of the universal scorefunction terms are doing things that have entropic contributions (the solvation term is sort-of the hydrophobic effect, which is partly-sort-of entropy).

The longer answer is, Rosetta energies are not physical energies, and deciding how to split them out requires far more data and time that we have so far. This is the sort of thing the Rosetta developers occasionally argue about (without ever answering the question) during late-night conversations at meetings.

Thu, 2014-12-04 13:55

HI smlewis,
Thank you very much. I can feel the difficulty in explicitly eluciding the entropy. If I understand correctly, the contributions in the scorefunctions are more of chemical energy.

Yours sincerely

Fri, 2014-12-05 07:39

In some respects, the Rosetta energy/score function is eminently practical - it is whatever it needs to be to get results to match up with what we want them to be. That is, the Rosetta energy function is set up such that folding simulations give output structures that look like native structures, that docking simulations give bound conformations which looks like native complexes, that mutational studies give mutational results that look like experimental results and naturally occurring proteins. Since these experimental results are based on the minima of the (Gibbs) free energy landscape, it tends to work out that the Rosetta energy function has minima that match (very crudely) those free energy minima.

This doesn't have great philosophical/theoretical support, however. Rosetta energies are always calculated based on single conformations, so that doesn't play well with the standard statistical mechanics definitions with entropy being from a consideration of ensembles. Rosetta also has several energy terms (fa_atr, fa_rep, fa_elec) which come directly from physics-based internal energy calculations, and would be considered purely enthalpic. On the other hand, Rosetta does include implicit solvent modeling (fa_sol) and a fair number of statistical potentials (e.g. fa_dun), which consider ensembles implicitly (the energy for the representative state is to some extent integrated over all the states which are close to that representative state), and thus would have an entropic contribution. (It would incorrect to assign them *entirely* to enthalpy/internal energy.) - That's all very nebulous, though, and any statistical mechanics professor would grade you harshly if you tried to argue the case in an actual thermodynamics class. As Steven says, it's something that gets debated heavily at late-night conference sessions without satisfactory conclusions.

As unsatisfying as it may be, the best answer is to dispense with attempting to assign Rosetta to enthalpy/entropy/free energy. The important question is: does the Rosetta energy function give you useful models, and does it match with experimental results? I'm unaware of anyone who has compared experimental energy/entropy/free energy values for various systems to Rosetta predictions of the same. Ultimately, that's what would need to happen to answer the question.

Mon, 2014-12-08 10:22

Hi R Moretti,
Thank you so much for spending time explaining the issues. I really appreciate it. I can strongly feel the endeavor involved to make REU closer to dG.

Yours sincerely

Fri, 2014-12-12 10:27