AMBER Archive (2002)

Subject: RE: Questions on crosslinking residues

From: Nicholson, James D Mr ARO (James.Nicholson_at_apg.amedd.army.mil)
Date: Wed Oct 02 2002 - 20:03:47 CDT

On Tue, Oct 01, 2002, Nicholson, James D Mr ARO wrote:

> I am modelling DNA with a cross-linking agent...
>
> What I did previously was to create a single residue which contained both
> crosslinked residues and the linker. When I imported this into leap, leap
> would correctly (for the most part) assign the atom types, but, I would
> still need to manually repair missed bonds. Apparently, it doesn't much
> like having a main chain that travels through the crosslink and out into
the
> other residue.
>

Imagine a double helix which contains two guanines that have been
crosslinked with an N7 alkylating agent. When you read in the pdb file, you
would expect to have two chains of DNA with a covalent bond between them via
the linker. But what happens with xleap is that, instead of coupling O5's
to the P from the preceeding residue and O3's to following residues on both
chains, you end up with a complicated mess where one chain of the double
helix has a link from O3' to P which skips the cross-linked residues
entirely and the other chain has two connections for each O5'-P and O3'-P,
one bond from each P to both of the O5' and O3' atoms.

When leap reads a pdb file, it assigns amber atom types, charges, etc. by
looking at the all_nucleic.lib, or whatever else you tell it to load with
your leaprc. In my case, I used leap to create an OFF file with a new
residue containing the crosslinked residues, and, I loaded it in my leaprc
at startup.

When I read in the entire double stranded DNA, leap tries to decide how to
hook things up, especially since my pdb file doesn't have CONNECT cards,
and, I'm not sure that leap would look at those even if they were there.

Because I collapsed two guanine residues into one crosslinked residue, I was
forced to rename half of the atoms (since a single residue must have unique
atom names). That's why I find this behavior so weird; one O3 is named O3'
and the other O3'5, yet both are connected to a P from the following residue
on one of the strands. So, the erronious bonds are probably not based on
the atom name.

I attribute this sort of behaviour to the main chain/side chain paradigm
which I see explicitly in the prmtop file and which I assume is used
somewhere in the residue templates. This particular application has two
"main chains" going through a single residue. And I think that's the
problem.

I am wondering if the residue templates can actually handle crosslinked DNA,
and, I'm curious how sander handles disulfide links, since that poses the
same problem. This is why I need to make manual changes to the structure.

I get leap to generate a coordinate and topology file for sander via manual
editing in leap, but, life would be easier if someone had a better way to do
this.

-Jim N.