You are here


5 posts / 0 new
Last post

Can other solvation models be implemented in the score function such as Poisson–Boltzmann or Generalized-Born?

Post Situation: 
Wed, 2011-08-31 17:14

I assume you mean, "are they already implemented". There is a class "GenBornEnergy" and a class "PoissonBoltzmannEnergy", so it looks like both are. The name for the method weights (in a scorefunction weights file) appears to be "gb_elec" and "PB_elec", respectively.

They are likely to be very slow compared to the default solvation model. I'm reasonably certain both are in good working order but I've never tried them.

Fri, 2011-09-02 09:16

Another thing to keep in mind is that, because of the need for short-range, pairwise decomposable energy functions for efficient use in most of Rosetta's protocols, the "Generalized Born" and/or the "Poisson Boltzmann" implementations in Rosetta may or may not match (or even be close to) standard GB/PB implementations in, for example, conventional Molecular Dynamics programs. Caveat experimentor.

Fri, 2011-09-02 12:08

I think our GB and PB are reasonably close to correct - I think they're just not pairwise decomposable, and thus can't be used in packing and are very slow. I've asked Yifan for comment - I'm told he's working on making them work with packing. (Or, given geography, is that what Yifan told you?)

Fri, 2011-09-02 13:01

Yifan says:

"There isn't a PB implementation in Rosetta. I put something in to call APBS externally and then read the potential file in. But it's not used generally for scoring.
As for GB, last I heard, there might still be a bug in it. Jim Havranek would have a good idea of what the status is. Also the GB scoring term is only for the pairwise interaction. I don't think the solvation term is implemented, but I could be wrong.

Mon, 2011-09-05 17:19