Re: amber-developers: Question concerning wildcards in dihedral definitions.

From: Thomas E. Cheatham, III <>
Date: Tue, 28 Jun 2005 17:23:47 -0700

> file. Should we perhaps do away with the wild cards all together to
> these ambiguities???

I agree; (can we get rid of LEaP too while we are at it?). What a mess.

I do not know the definitive answer here (are Piotr or Wendy on this
list?), but my recollection which is likely flawed is that if a specific
torsion is present, it overrides the general (wildcards), independent of
periodicity... To definitively check this (i.e. what the force fields
were originally developed off of), we could run your files through PLEP
(prep-edit-link-parm). As I've got compiled versions of these old codes,
I in fact just did this...

Amazingly PLEP still works on one of my SGI's. On nuc.pdb, I get (for
dihedrals containing O4' AND H5'?)

           peak phase periodicity
       6: 0.156 0.00 3.0 :1.H5'1 :1.C5' :1.C4' :1.O4' (6,5,8,10)
       9: 0.156 0.00 3.0 :1.H5'2 :1.C5' :1.C4' :1.O4' (7,5,8,10)

Unfortunately, this is different than what I get both from your first dna
prmtop (dna.prmtop)

      38: 0.250 0.00 1.0 :1.H5'2 :1.C5' :1.C4' :1.O4' (7,5,8,10)
      41: 0.250 0.00 1.0 :1.H5'1 :1.C5' :1.C4' :1.O4' (6,5,8,10)

and your second prmtop (dna2.prmtop):

       6: 0.250 0.00 1.0 :1.O4' :1.C4' :1.C5' :1.H5'1 (10,8,5,6)
E 7: 0.156 0.00 3.0 :1.O4' :1.C4' :1.C5' :1.H5'1 (10,8,5,6)
       8: 0.250 0.00 1.0 :1.O4' :1.C4' :1.C5' :1.H5'2 (10,8,5,7)
E 9: 0.156 0.00 3.0 :1.O4' :1.C4' :1.C5' :1.H5'2 (10,8,5,7)

The only good news is that LEaP in the second case at least has the brains
to ignore end-group (1-4) interactions on the second specification of the
torsion. The bad news is that all three of these methods to
create the prmtop disagree.

I am going to rebuild the prmtops w/ LEaP and parm94 to verify that I can
see what Ross does (as I did parm94 with PLEP not parm99). I will also
check to see if it relates to end group interactions (terminal residues).
However this scares me quite a bit, especially since we are seeing
significant DNA force field problems in many longer simulations and that
some of these weird differences are showing up in chi related parameters
(not just dihedrals with hydrogen).

However, as an aside I do know that I was able to get CHARMM and AMBER to
produce identical energies/forces with a version of Cornell et al. at one
point and need to check if my AMBER prmtop's were PLEP'ed or LEaP'ed at
that point by trying it all again.

Ross, please let me know if you get further in tracking things.

Received on Wed Apr 05 2006 - 23:49:55 PDT
Custom Search