On Tue, 2014-05-06 at 10:04 -0700, Sidney Elmer wrote:
> Hi all,
>
> I am a little nervous about the dihedral terms I just found in one of my
> prmtop files for a system with DNA (built with the amberff99SB-ILDN force
> field). Below I show all the dihedrals for a CYT residue (I haven't
> checked if this problem is occurring in other residues or in other force
> fields). Note the entries that I have marked '#' or '##' or '###', these
> are the terms I am questioning. From what I understand, any term that
> begins with 'E' extends the definition of the dihedral term preceding it.
> Is that correct? If that's the case, then it would only make sense that
> the atoms which define the term would have to be exactly the same in each
> row. Is that correct?
>
> If both of my assertions above are correct, then there is a big problem
> because the dihedrals I have marked '#' or '##' or '###' do not follow
> those rules. Is this a bug in tleap?
>
> If on the other hand, these entries below are allowed, could someone please
> explain to me what they mean, because I am confused if this is designed
> this way. Thank you for your help.
I'll try to keep this short-ish, but here's some background to start.
Amber defines both impropers and proper dihedrals in the same list and
uses the same functional form to treat them. In the prmtop, improper
dihedrals are indicated with a negative 4th term, whereas
end-group-ignored dihedrals have a negative 3rd atom index (see
http://ambermd.org/formats.html). I think rdparm uses "E" to denote
ignored _E_nd groups and B to denote that _B_oth end-groups are ignored
AND it is an improper, but don't quote me on the notation there).
The dihedral list is used in the Amber source code as an effective 1-4
nonbonded pairlist and the 1-4 atom pairs are actually excluded from
each other in the nonbonded list. Therefore, the prmtop has to indicate
somewhere in the dihedral list which atoms need to have their 1-4
interactions computed in the 1-4 nonbonded list and which ones need to
be omitted.
As you pointed out, multiterm dihedrals (in which all 4 atoms are
exactly the same) should have at most 1 of their dihedrals have a
positive 3rd atom index so that you don't calculate the same 1-4
nonbonded interactions multiple times. For proteins, most of the
negative indexes mean precisely that -- you have a multiterm dihedral
and only want to compute the end-group interactions once.
A topological complication arises from rings (6-membered rings and
smaller). Consider the cartoon below (hopefully you have a fixed-width
font to view it in):
2----3
/ \
1 4
\ /
5-----6
Atoms 1 and 4 are partners in two dihedrals -- 1-2-3-4 and 1-5-6-4. You
only want to count those end-groups once, so ONE of those dihedrals
should be marked with an "E" even though it's not a multi-term dihedral.
Consider 5-membered rings
2----3
/ \
1 4
\ /
\_ _/
5
Now atoms 1 and 4 are dihedral partners in the 1-2-3-4 torsion, but
they're also connected by an angle via 1-5-4 and should be excluded
entirely, so again 1-2-3-4 should be marked "E" even if there is only 1
term because those atoms are excluded on the basis of being a 1-3 pair.
Rings are far more prevalent in nucleic acids than amino acids
(obviously), but you will see the same kind of thing with phenylalanine
or tyrosine, for instance. Note, any ring larger than 7 atoms will not
have two atoms connected by multiple dihedrals, angles, or bonds (unlike
6-, 5-, and 4-membered rings), so only these 3 ring sizes are
special-cased.
Hope this helped,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue May 06 2014 - 11:00:05 PDT
