[gmx-users] Another dihedral angles question

[David Cerutti]

Hello,

Many thanks to Erik Lindahl and David van der Spoel who have tried to help
me with reverse-engineering the dihedral angles potentials.  However, I'm
still not understanding something with this.

In ffgmxbon.itp, there are two subsections relating the proper dihedrals.
The first deals with "i" and "l" atoms, and I don't know what the q0 and
cq parameters refer to.  The first looks like a potential (I note that it
is zero in most cases) and the second looks like a phase angle.  However,
I'm confused.  The list of i and l atoms doesn't look like it could
possibly be exhaustive, and I really don't know what q0 and cq are.

The second subsection is more clear to me now, and I think I'm correctly
using phi0 as the amplitude, cp as the phase angle, and mult as the
multiplicity to multiply a given dihedral angle by inside the cosine
argument.  However, I'm then confused a bit by some of the lines that
have been commented out.  Dihedrals involving CS1 and CS1, for example,
could apparently have three potentials associated with them, but the last
two are degenerate (why not simply combine all the potentials that have
the same phase angle?  Is it intended that only one of the potentials is 
to be used in a particular instance?).

Finally, I need to make sure I am not overcounting things.  If I see a
structure like

     O1
     ||
N    C--N2
 \  /     \
  CA
  |
  CH3

(Hope you've got a courier font)

I see dihedral angle potentials that need to be applied between O1 and the 
CH3 united atom, N2 and the CH3 united atom, 01 and N, and N2 and N.  So,
I'm going to compute the four dihedral angles I just described and plug
them into (for the j,k atoms C and CA)

   potential = kphi(C:CA)*cos(mult(C:CA)*angle-phi0(C:CA))

and then I'll have to do something with the four pairs of i:l atoms, but
I'm not sure how to use q0 and cq in that case even if parameters for
these values are given for O1:CH3, N:N2, N2:CH3, and N:O1.

Any help will probably, finally, solve my conundrum.  The docking prgogram
I'm writing works pretty well as is, but I think that thhis touch-up to
the potential function will only improve things.

Thanks as always,

Dave