On Wed, Dec 17, 2014 at 2:20 PM, Hallel Freedman <hfreedma.ualberta.ca>
wrote:
>
> Hi Jason,
> OK now I have some things to try. Thank you very much for your
> suggestions.
> Regarding the dielectric boundary, because a set of nucleotides shares so
> much similarity and because of the spacing between the ribose and phosphate
> atoms by the c5' group, if it were technically possible to simply forget
> this contribution to the dielectric boundary from neighbouring phosphate
> atoms, and perform the main sum of energy contributions over only
> non-phosphate atoms, I believe that this would produce valuable relative
> binding free energy data. However I understand that within your
> implementation, this may be very difficult to achieve.
>
This isn't really an implementation detail. It's a property of solving
the PB and GB equations themselves. They both depend on the physical shape
of the molecule, and the shapes with and without the phosphate groups are
different. If the phosphate was in exactly the same place for every
ligand, then you will probably get a lot of the error to cancel when
rank-ordering the different ligands, but I'm not sure how sensitive the PB
energies will be to small changes in phosphate position and orientation.
Ignoring the contribution of a set of atoms to the shape of the molecule
(i.e., to the dielectric boundary) is a decision you have to make *before*
you do the calculation -- it's not something you can simply post-process
from the output regardless of the information that is printed. This is
equivalent to the practice of stripping the phosphate groups with ParmEd or
cpptraj (but then of course your net charge will be weird and probably not
an integer).
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Dec 17 2014 - 12:00:02 PST
