Menu

#1250 EquivalentClassPartitioner aromaticity bug

cdk-1.4.x
closed
nobody
5
2013-05-23
2012-08-28
No

Hi,

there is a bug in the EquivalentClassPartitioner because it is not able to assign equivalent classes properly for aromatic molecules. For example, the molecules 1H-isoindole (CHEBI:33178) and 2H-isoindole (CHEBI:33179) should give the same class partition because all atoms are aromatic. The current implementation sets first the class corresponding to a sequence of single and double bonds and only takes aromaticity into account as last resort. I think I know how to fix the bug and I'll try to submit a patch asap.

Cheers,
Luis

Related

Patches: #584

Discussion

  • Luis F. de Figueiredo

    I have uploaded the patch with some additional tests for bug 3562476. I corrected the bug by changing the adjacency matrix calculated by class ConnectionMatrix. Now, aromatic bonds have value 1.5 in the adjacency matrix which is the value suggested in the paper where this class is based on.

    Given the relevance of this distinction between bond orders in the adjacency matrix, one should actually consider changing the way this matrix is calculated in CDK but I don't have a clear vision of the impact of this change in other classes (but it's probably a big impact).

    Cheers,
    Luis

     
  • John May

    John May - 2013-05-23
    • status: open --> closed
     
  • John May

    John May - 2013-05-23

    Resolved