Hi Christoph, On Wed, May 19, 2010 at 10:07 AM, Christoph Steinbeck <steinb...@ebi.ac.uk> wrote: > As I said in my other email: > We may want to impose some "stability" on atom types and then go through > subsequent rounds of atomtyping for a molecule until they stabilize.
Indeed. The current unit tests actually tests this stability into one recursion. There is another issue, however: what is stability. I think the above refers to algorithmic stability. Tautomerism, instead, has to do with chemical stability: O=C-NH <-> HO-C=N These are two different chemical species, however, with a low energy barrier. It is actually this resonance what defines the N.amid atom type... as the four atoms are in a single plane, reflecting the double bond nature character of both the C-O and the C-N bond. If the input is the first, and the C-N part of a six ring with two further double bonds [0], the representation does not reflect an aromatic ring... the tautomer alcohol has a proper benzene ring... this is a matter of structure normalization, very much like how to represent a NO2 functional group (four bond equivalents on nitrogen and charged, or neutral nitrogen with five bond equivalents). Now, I'm hesitating in introducing normalization code in the atom typer... need to think more about it; more sitatuations to estimate the impact... e.g. if we introduce keto-enol tautomerism in the atom typing, how will we estimate the stable form? would that be 'stable' enough to apply to any tautomerism... An approach like InChI takes is quite interesting, but not really compatible with the current atom typing scheme.... C.sp2 changes into C.sp3, etc, etc... Egon 0.http://chempedia.com/names/2-1h-pyridon -- Post-doc @ Uppsala University Proteochemometrics / Bioclipse Group of Prof. Jarl Wikberg Homepage: http://egonw.github.com/ Blog: http://chem-bla-ics.blogspot.com/ PubList: http://www.citeulike.org/user/egonw/tag/papers ------------------------------------------------------------------------------ _______________________________________________ Cdk-user mailing list Cdk-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/cdk-user
