VMD-L Mailing List

From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Mon Jul 09 2007 - 21:14:31 CDT

Hi Irene,

I'm not sure what's causing the errors you encountered (not sure exactly
what *could*, besides something like having multiple definitions of the
same bond, or two atoms with the same name), since I've never really
seen this happen, but I can offer some general input.

The nomenclature you suggest from existing topologies is a good rule of
thumb, and will lead to well organized topologies; however, it is *not*
mandatory. NAMD and psfgen don't care what the atoms are called, as long
as every atom name in a residue is unique and they aren't too long, so
if you can make your life easier by keeping similar atoms in sialic acid
with the same names as those in NAG, you should by all means do so and
save some headaches.

Also, as a general rule it's almost always easiest to only define bonds
in your topology file and then specify "AUTO ANGLES DIHE" to
automatically generate angles and dihedrals from the bond connectivity.

The IC lines are *only* used for coordinate guessing, so if you can
provide a complete sialic acid residue (or at least one that is only
missing hydrogens), there's not really any reason to mess with them.
There isn't a need for any particular ordering in how you put these
entries in; you should, however, build back to atoms whose positions are
already known (as you inferred). IC assignment is still something of a
black art to me (not as simple as making a z matrix, for example), but
making sure that every atom has at least one bond, angle, and dihedral
(or improper) specified is probably a good way to go. I try to avoid
them when building top files if I know that I can provide most or all of
the coordinates as input, but this shouldn't be the root of the problems
you're having.

Hopefully this is helpful... I'd be happy to look at the top file and
see if I see anything that could be causing this.
Best,
Peter

Irene Newhouse wrote:
> I've successfully built a topology file for N-acetylglucosamine. Now I'm trying to do the same for sialic acid, which is chemically fairly similar to NAG, but is proving to be much harder.
>
> I'm running into a bizarre problem - psfgen works & doesn't print any unusual error messages. However, my psf file contains 2 angles which are of this form: CT2-HA-CT2, where CT2 actually refers to the same atom(!) and HA-CT2-HA, again, where HA is the same atom, not 2 different ones [I've been peeking at the psf file itself to get that info]. How did I stumble on the problem? When I try to run NAMD with this psf file, I get an error message stating that there are no parameters for an angle of type CT2-HA-CT2 [duuuh!] After messing about some, I started looking through the psf file & that's how I found the other error. That one would likely not produce a NAMD error message, wrong as it, too, is.
>
> Can someone supply me with some general guidelines for building a topology file which prevent psfgen from getting confused like this?
>
> So far, by generalizing from existing toplogies, I get the feeling that you start at point A & go along the backbone. Once the backbone is pinned down, so to speak, you go back for the side chains. You define your backbone to be as long as possible. It turns out that the standard numbering for sialic acid differs more from that for NAG than you'd think, based on the structure, because the 'maximal backbone' rule is applied in the numbering scheme as well. [If this interests you, we can email about this off-list]. In the ICs, your references should go 'back' toward atoms in the molecule you've already specified, by which I mean if you have an H7 on C7, then your IC should read H7 C7 x6 x5 where 'x' means the atoms in positions 6 & 5 of the backbone, C or not, as opposed to 'forward' toward atoms you haven't yet dealt with. [This seems to result in a fair amount of duplication of entries of previously-defined angles and bonds in IC statements].
>
> A major source of confusion for me is when do you decide proper dihedral definitions are not sufficient, and you need to add improper dihedrals? Also, how many specifications are enough? For instance, for the ring atoms in sugar molecules, the CSFF topology files include 2 instances of improper dihedrals in which a given atom is the central atom, but that's not the case for straight-chain amino acids. I'm finding that if I over-specify H's on the 2 last carbons of the 'backbone' of sialic acid, then psfgen complains that the H on the 3rd-to-last C of the 'backbone' has to be guessed poorly - somewhat non-local, no? And not quite obvious why that should be the case to this novice. But none of the changes which remove that complaint result in disappearance of the CT2-HA-CT2 angle from the psf file.
>
> The atom in question is like the CH2OH group in serine, and yes, I modeled my topology on serine, and that didn't help.
>
> I'm willing to send out my files if that proves necessary, but I'm hoping that someone can give me some additional guidelines I've missed that allow me to dig myself out.
>
> Thanks!
> Irene Newhouse
>
>
>
>
> _________________________________________________________________
> Missed the show? Watch videos of the Live Earth Concert on MSN.
> http://liveearth.msn.com
>