[gmx-users] Re: Updated atom name nomenclature table, pre-release available

[David Mathog]

David van der Spoel wrote:

> David Mathog wrote:

> > 1. It wasn't at all clear to me how to assign HB1 vs HB2,
> > HD1 vs HD2 and so forth for Amber.  Or if it even mattered if those
> > were reversed.
> > 
> It does matter in case it is used for e.g. NMR refinement and they are 
> different (e.g. not on ALA but everywhere else for the HB*).

Ok.  I'm no chemist, could somebody who is please tell me which
way it goes for the Amber column?

> 
> > 2. From spelunking the gromacs code it seems that Gromacs uses "O"
> > for both "O" and "OXT" at the C terminus.  So the two "O" entries
> > in that section of this file are NOT a typo.  Conversely, this
> > part of Gromacs really, really, really should be changed since having
> > two different atoms with the exact same name is bad form.
> But in general these are renamed to O1 and O2, aren't they?

Eventually, but only for Gromacs FF.  The tricky thing about
pdb2gmx is that it has residue and atomic name changes scattered
over multiple function levels.  Ideally it should all be in one
place.  The initial OXT->O conversion is somewhere below read_pdball().
I cut in after that, at process_chain() to add name translation
PDB->FF (most of them seemed to be in that routine anyway), and at
that point those atoms are still O,O.  To get it to work with
both Amber and Gromacs O,O is left alone for Gromacs but translated
to OC1,OC2 for Amber. Because there are two same named atoms doing
so requires a scan over all atoms in the residue to count "O" atoms.
About 30 lines after after process_chain() near the
" Set termini " comment the names are finally changed from
O,O to O1,O2. But that only works for Gromacs because it has a
tdp file.  In general I think it creates program structure issues
that OXT is translated to O so early. Honestly I don't see why
it was ever done that way since the code to convert O,OXT -> O1,O2
is much simpler than what is there now. 

> I presume you only intend to do that in this file? 

Right.

> How about giving 
> every amino-acid residue also an H1 H2 and H3 in addition to H?
> As long as it is for name translation only I don't see a problem with it.

That was the idea.  It's purely for name translation.


> Do I understand correctly that you are implementing code that translates 
> one naming scheme into another?

Not N formats to N formats directly, just N->PDB and PDB->N.

> How about some heuristic code that derives from a complete pdb 
> file what is the most likely source? How about pdb files that use 
> multiple standards? Not meaning to discourage you...

Walk first, then run.

Regards,

David Mathog
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech