[Cdk-devel] Bucky ball with inconsistent aromaticity detection

[Mark R. <ma...@eb...>]

Hi all,

for OrChem, I've been testing fingerprinting and substructure search. 
Part of the fingerprinted aspects are to do with double bonds and 
aromaticity.
The so called complex bucky balls (see a mol file attached) kept giving 
me inconsistent results. I'm working with cdk 1.2.


I have attached a mol file, its picture and a Java program (find them at 
http://sourceforge.net/mailarchive/forum.php?forum_name=cdk-devel)

The program reads the mol file, creates a compound and detects 
aromaticity. It does this in a loop, it repeats xx times exactly the 
same thing.

When I run this (I hope you get this behaviour too in your JVM) I see to 
my horror the following:


arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 55  double 3 <<
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 55  double 3 <<
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2
arom 56  double 2


So at any unpredictable time, it can count 55 instead of 56 aromatic 
bonds. It seems completely random. Screws up my fingerprinter.
My question is if I'm doing something obviously wrong ? If not, I'll 
have a look further at the underlying CDK classes to find a possible bug.


Mark.


PS apologies in advance : I'll be off on holidays soon so might not be 
replying for a while.