I indexed several hundred structures from a file containing smiles with fastsearch-index using OpenBabel 2.3.1 built from source in Slackware Linux. Some of the structures contained azides in the form of: [N-]=[N+]=N
When Im trying to search them later the fast search index seems to not work properly. I can
t find azides by structure. Other structures seems to be OK.
The problem is a broken method to recognize both forms of dative bonds, e.g. -N=[N+]=[N-] and -N=N#N, as the same. I have now committed to SVN (r5042) some mods to make this work. The datafile and the target molecules can now be in either format: all the chemically identical speies appear the same. This also applies to the bug report #840.
A longer article with more details is attached.
Is there a chance for sending me a patch so I could compile myself the latest 2.3.2 ?
Last edit: Witek Mozga 2012-12-01
OK, I managed to get your revision from subversion myself and search for azides works OK now, however when i`m trying to search similarities against index with
babel index.fs results.smi -sCN -at5
I got segmentation fault. It worked before.
On my Windows system I don't have any difficulty with the similarity search. Could you try
rebuilding and remaking the fs file? An indication from a debugger of where it is crashing would be useful.
It works. The reason was outdated documentation on the openbabel website, namely the site http://openbabel.org/wiki/Tutorial:Fingerprints says that this is sufficient:
babel mymols.fs results.sdf -sN#Cc1ccccc1C#N -at0.6
and it in fact was sufficient in 2.3.2, but now it looks that every use of index require -ifs option like this:
babel -ifs mymols.fs results.sdf -sN#Cc1ccccc1C#N -at0.6
then it works. The documentation needs to be fixed. I
ll try to register to the wiki and fix it in this moment. EDIT: Unfortunately I
m unable to change the docs in the wikiLast edit: Witek Mozga 2012-12-02