Fingerprint was ignoring pseudo atoms... which these entire fragments… #305
Conversation
…gerprint on an anon graph only has 7 bits set.
|
Grrr hold off on this... the hybridisation fingerprint was doing more than I thought. If you think about it in this case it really is useless as you should get n paths. |
|
Okay to commit now, the last commit shows how using the fingerprint isn't useful - all the FPs are the same :-). |
| @@ -365,8 +365,9 @@ public void makeCanonicalSmileFromRingSystems(String dataFileIn, String dataFile | |||
| public List<IBitFingerprint> makeFingerprintsFromSdf(boolean anyAtom, boolean anyAtomAnyBond, | |||
| Map<String, Integer> timings, BufferedReader fin, int limit) throws Exception { | |||
|
|
|||
| IFingerprinter fingerPrinter = new HybridizationFingerprinter(HybridizationFingerprinter.DEFAULT_SIZE, | |||
| HybridizationFingerprinter fingerPrinter = new HybridizationFingerprinter(HybridizationFingerprinter.DEFAULT_SIZE, | |||
There was a problem hiding this comment.
The basic idea of this one is that sp2 hybridization is a lot more stable than aromaticity...
There was a problem hiding this comment.
Yep... but the atoms were all anonymised! :-)
There was a problem hiding this comment.
But only in the 3D builder stack, right? There this was used as a backup: first it would try to find a matching ring system with the correct elements, and second, if no such ring was found, at least a ring system for the same ring sizes but possibly different chem. elements in the ring...
There was a problem hiding this comment.
(But the general sp2 FP should not anonymize atoms by default!
There was a problem hiding this comment.
No it doesn't... but yes only the template matcher;
But only in the 3D builder stack, right? There this was used as a backup: first it would try to find a matching ring system with the correct elements, and second, if no such ring was found, at least a ring system for the same ring sizes but possibly different chem. elements in the ring...
Maybe that was intended but it wasn't doing that.
There was a problem hiding this comment.
Mmm... that's a serious regression then...
… were.
Fix test regressions, not sure why I missed this.