[Joelib-devel] Atom types and protonation models

[Joerg W. <we...@in...>]

Hi all,

i was adressed recently with the following question:
Is there any standard for atom types and a protonation
model available?
Or from a more practical standpoint of view. Is there
any multiple molecule file available with a standard
definition of atom types AND protonation models
for one (or different pH) value(s) ?

I did'nt know a real good forum for this question, so
i wanna adress this problem to your community of
developers. Because this is a posting to
different forums, i would recommend the Openbabel
discussion list for a response (if this is o.k. !),
otherwise you can send responses to the JOELib or CDK
(if this is o.k. !) developer mailing list.

If you aren't interested in this problem, you can stop
reading here, because i will explain the problem in
more detail, from my standpoint of view.

------------------------------------------------------

Atom types are required for the following tasks:
- file conversion, e.g. SDF->Sybyl mol2
- protonation models which are BASED on this previous
  atom type assignments
- descriptor calculation methods which BASED on this
  previous atom type assignments

There are a lot of descriptor calculation programs
available and i think that every program uses it's own
atom typer facility. Additionaly this causes different
results for different programs !
From the computer scientist point of view that's absolutely
awful !!!
1. An algorithm produces always the same result !
   That's o.k.
2. Models, like atom typer models and protonation models
   are based on expert knowledge, BUT it would be
   more efficient AND transparent if there will be a
   set of molecules available WITH assigned atom types.
   Any hypothetically 'new' atom typer can be
   cross-checked against this multiple definition file.
   If it fails on single atoms, the atom type assigning
   process (e.g. based on SMARTS, as in Openbabel/JOELib)
   can be corrected and any side effects changing this
   definition can be double checked again and so on.

I used the Openbabel definitions and corrected
one or two entries, but i didn't check all definitions.
That will be hard work, and all previous developers of
Babel, OELib, Openbabel have invested a lot of time to
develop these atom types and definitions.

I'm busy with other main task, but nonetheless it
would be really usefull:
1. to have a standard or some standard files.
2. if any wrongly assigned atom type will be send
   to a mailing list, where these files should be stored
   in a transparent and usefull way.

Now the problem:
I think this problem is to big for a handfull of developers:
- without knowing the standard
- without knowing how to represent the standard
- without knowing who will be able to collect the
  definition files
- without having a transparent method to debug
  the atom type assigning process

Solution suggestions:
1. I would suggest a special mailing list in the
   OpenBabel or JOElib project to adress only 
   atom type and protonation model assignments.
   
2. The assignemts should be defined in a transparent
   and reproducable way. Any suggestion? Which
   are practical ? Technical docu ?
   
3. I think it will be definitely possible to extend
   Openbabel and JOELib to have a transparent debugging
   mode for the atom type assignments.

That's an open discussion:  Feel free for some
constructive comments and fruitful suggestions.

Regards, Joerg

Dipl. Chem. Joerg K. Wegner
Univ. Tuebingen, Computer Architecture, Sand 1, D-72076 Tuebingen, Germany
Tel. (+49/0) 7071 29 78970, Fax (+49/0) 7071 29 5091
E-Mail: mailto:we...@in...
WWW:    http://www-ra.informatik.uni-tuebingen.de