Re: [Cdk-devel] aromaticity detection failing

[Christoph S. <er...@do...>]

Rajarshi Guha wrote:
> 
> On Jun 25, 2007, at 10:22 AM, Christoph Steinbeck wrote:
> 
>> Parsing such a smiles into a CDK atomcontainer will (at best) set the
>> aromaticity flag of the atoms and bonds to "true".
>> The HueckelAromaticityDetect, however, just looks at the double bond
>> pattern and ignores aromaticity flags.
>> Looking at the name of the class, one has to admit that it does what the
>> name implies. :-)
>> I'd say, however, that this behaviour of ignoring the aroma flags should
>> be documented.
> 
> Hmm, in that case the SmilesParser seems to be the source of the problem
> as it assigns single bonds to the whole of the benzene molecule

I think this comes down to the old CDK discussion on how "intelligent"
we want our algorithms to be.
My preference always was the algorithm should do the (documented!!!)
minimum and leave the user the freedom to apply other, additional
methods to achieve he or her goal.
In this case, the user would need to run a kekule-izer over the parsed
SMILES to get all kekule structures in agreement with the input.
The input did not give any information here, so we should not do deliver
any magic results to the user which were not in the input.

In my opinion, the Aromaticity Detector should return true if all atoms
in a ring have the aromaticity flag set to true.

Cheers,

Chris

-- 
PD Dr. Christoph Steinbeck
Lecturer in Chemoinformatics
Univ. Tuebingen, WSI-RA, Sand 1, D-72076 Tuebingen, Germany
Phone: (+49/0) 7071-29-78978   Fax: (+49/0) 7071-29-5091

What is man but that lofty spirit - that sense of enterprise.
... Kirk, "I, Mudd," stardate 4513.3..