Camelia, the Perl 6 bug

IRC log for #cdk, 2010-06-23

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
04:08 jpansane1 joined #cdk
04:45 egonw joined #cdk
05:31 sneumann__ joined #cdk
06:12 jbrefort joined #cdk
06:38 sneumann__ joined #cdk
06:56 egonw joined #cdk
06:56 jbrefort joined #cdk
07:01 jbrefort joined #cdk
07:05 Gpox joined #cdk
07:21 maclean joined #cdk
07:29 carsten joined #cdk
07:56 maclean joined #cdk
10:57 maclean joined #cdk
10:57 maclean hi
10:57 zarah oh hai maclean
10:58 egonw hi maclean
10:58 maclean hi egonw
10:58 maclean how' it going?
10:58 maclean +s
10:59 egonw ah, well, you know...
10:59 egonw too much to do :)
10:59 maclean No! You? Surely.
10:59 maclean :)
11:00 maclean Well, when you have time, it would be good to talk about 160-cip.
11:21 s9asad joined #cdk
11:43 maclean Hi s9asad
11:43 s9asad Hi maclean
11:44 maclean Please do test the signatures on whatever examples you have
11:44 s9asad well I am exited to test it
11:44 s9asad is this now available with CDK?
11:44 maclean I expect that structures that test the SMSD probably will exercise signatures too
11:45 maclean It will be soon. Rajarshi reviewed it, and all the javadoc and unit tests are (finally) in place.
11:45 egonw s9asad: that patch is approved, and I will merge it in this afternoon
11:46 s9asad @maclean, yes I can test it with SMSD based examples too
11:46 s9asad @Egon great
11:46 s9asad @egon, could please review my SMSD patch, when you have time
11:47 s9asad I was wondering if signature can be used like CDK fingerprint?
11:47 maclean Since a canonical signature is something like a graph certificate, it can be used for a limited test for isomorphism.
11:47 maclean Fingerprint? Hmmm.
11:48 egonw s9asad: see my blog post -> http://chem-bla-ics.blogspot.com/2010/​06/qspr-modeling-with-signatures.html
11:48 zarah egonw's link is also http://tinyurl.com/3a6yblx
11:48 egonw s9asad: it's on my list, but really busy... no promises
11:48 maclean I suppose that I don't know what a fingerprint is essentially.
11:48 maclean It's something like a hash, I suppose.
11:48 egonw fingerprint: bit string that reflects a molecular structure
11:48 s9asad yes, it's a hash
11:49 s9asad yes but you could load it with weighted fingerprints
11:49 egonw but hashing is not necessarily involved
11:50 maclean "A hash function is any http://en.wikipedia.org/wiki/Algorithm or http://en.wikipedia.org/wi​ki/Function_(mathematics) that converts a large, possibly variable-sized amount of data into a small datum"
11:50 zarah maclean's link is also http://tinyurl.com/48kp7
11:50 maclean gah.
11:50 maclean sorry.
11:50 s9asad yes.. as egon pointed out
11:51 s9asad if two molecules share some siganture (partly), then it might also share some supgraph...
11:51 maclean True.
11:51 s9asad if this is true then one could transform signatures to fingerprints
11:52 egonw yes, see the paper from FAulon in 2003
11:52 maclean The problem being that you have to have equivalent heights.
11:52 egonw this is what I did on my blog post of saturday
11:52 egonw maclean: indeed
11:52 egonw maclean: same for CDK fingerprint
11:52 egonw you pick a height there too
11:52 maclean Hmmm.
11:53 s9asad can you trace back (reconstruct) the mol with your signature?
11:53 maclean yes
11:53 s9asad thats why you need a height
11:54 s9asad if you define a default height /optimal, then you can have a fp's, which can be used to compare mol but not necessary reconstruct them.
11:54 s9asad does this make sense or I am off track
11:54 maclean IMolecule reconstructed = MoleculeSignature.fromSignatureString("[C]([C])", NoNotificationBuilder.getInstance())
11:55 egonw maclean: but it only works for molecular signatures, irght
11:55 egonw ?
11:55 egonw not from atomic signatures
11:55 egonw unless the have a depth equals to the graph diameter of the molecule
11:55 maclean The distinction is hazy in this case.
11:56 maclean If the height is less than D, you get a substructure.
11:56 egonw right
11:56 s9asad yes
11:56 maclean s9asad : I suppose.
11:56 egonw so, if the height=D, a single atomic signature will do
11:56 maclean Yes
11:56 s9asad so lets say height of 7/8 should be good for most of the small mol
11:57 s9asad if they are larger we can capture the suggraphs as fps
11:57 s9asad subgraphs
11:57 maclean Hmmm. I suppose. What would this be for?
11:58 maclean Well, fingerprinting, obviously, but...
12:00 s9asad if we have fps of mols using signature then, mol comparison might be more precise/sensitive and faster than subgraph comparison
12:00 maclean Ohhh. So /that's/ what geometric hashing is. (thanks wikipedia :)
12:01 s9asad :-)
12:02 s9asad so you got the clue !
12:02 maclean Using a certificate for comparison is faster if making it is reasonably quick, and it is then reused multiple times, yes.
12:03 maclean Like if you convert a list of structures to (canonical) smiles, and then compare all-v-all.
12:03 maclean s9asad : I do, yes.
12:03 maclean I think.
12:05 maclean Oh, and one not of warning : there are structures that take a long (O(n!)) time - where a number of rings are connected to a central spoke atom.
12:10 s9asad yes but can't they be stored as signature like smiles and retrieved as FPs when need unlike graphs.
12:11 s9asad left #cdk
12:42 bag joined #cdk
13:26 CIA-47 org.openscience.cdk: Arvid Berg bioclipse2.4 * r167989e / plugins/org.openscience.cdk.contro​lbasic/src/org/openscience/cdk/con​troller/AlterBondStereoModule.java :
13:26 CIA-47 org.openscience.cdk: Enables AlterBondStereoModule to cycle between different stereo modes
13:26 CIA-47 org.openscience.cdk: UP->UP_INVERTED->NONE->UP...
13:26 CIA-47 org.openscience.cdk: DOWN->DOWN_INVERTED->NONE->DOWN... - http://bit.ly/bQ0Hm8
16:10 CIA-47 joined #cdk
16:15 sneumann joined #cdk
16:58 egonw joined #cdk
17:28 egonw joined #cdk
18:38 sneumann_ joined #cdk
20:30 jbrefort joined #cdk
20:35 CIA-47 cdk: maclean master * r935b5f6 / src/main/org/openscience/cdk/​signature/AtomSignature.java : Convenience constructors for AtomSignature that take IAtoms rather than atom indices (+39 more commits...) - http://bit.ly/d6l5uf
20:56 egonw_ joined #cdk
21:01 CIA-47 cdk: annulen * r15552 /jchempaint/trunk/src/main/org/openscience/jchem​paint/dialog/editor/PropertiesModelEditor.java: Removed obsolete line
21:12 s9asad joined #cdk
21:31 s9asad joined #cdk

| Channels | #cdk index | Today | | Search | Google Search | Plain-Text | summary