Camelia, the Perl 6 bug

IRC log for #cdk, 2011-09-22

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

All times shown according to UTC.

Time Nick Message
02:05 sneumann_ joined #cdk
03:57 egonw joined #cdk
05:24 sneumann_ joined #cdk
06:53 egonw joined #cdk
07:02 Gpox joined #cdk
07:53 jonalv joined #cdk
08:10 jbrefort joined #cdk
11:38 maclean joined #cdk
12:05 jonalv It might be that I am silly but I could use some help. How do I find out how to convert my SD string into an atomcontainer using CDK?
12:27 maclean Sorry, jonalv, was reading wikipedia, and not paying attention to the chat :)
12:27 jonalv nw
12:27 maclean SD is part of an SDF, right?
12:28 jonalv well what would you call it?
12:28 maclean a molfile, I think.
12:28 jonalv okey
12:28 jonalv a molfile string then
12:28 maclean Ah, ok, then you can make a java.io.StringReader and pass that to an MDLV2000Reader or similar.
12:29 maclean So, the string reader reads from the string, and the molfile reader interprets.
12:29 jonalv sounds complicated :)
12:29 maclean Heh, a little. I guess it would be nice if readers did this themselves.
12:29 jbrefort joined #cdk
12:29 jonalv I wish I had the Bioclipse cdk wrapping
12:31 jonalv how do I get the atomcontainer from that reader then?
12:31 maclean IAtomContainer atomContainer = mdlReader.read(new AtomContainer());
12:32 jonalv oh
12:32 maclean Or some other implementation, like NNAtomContainer.
12:32 jonalv I give it an atomconatiner
12:32 jonalv aha
12:32 maclean Yes, a bit counter-intuitive, but it makes the process implementation independent, I think (wasn't my idea! :)
12:35 maclean jonalv: What are you using it for, anyway?
12:35 jonalv just need a test mol
12:35 jonalv don't wanna mess with eclipse filapths to read in the entire file
12:35 maclean heh.
12:36 jonalv yey onw line
12:36 jonalv IAtomContainer testAC = new MDLV2000Reader(new StringReader(testmol)).read(new AtomContainer());
12:36 jonalv maclean:  thanks
12:36 maclean no problem :)
12:36 jonalv now to see if this works :)
12:37 maclean dot dot dot
12:39 jonalv hm I clearly did not understand the code I was trying to test with this
12:40 jonalv but  I have no reason to believe that the molfile parsing is to blame
12:40 maclean Oh?
12:40 maclean (Bug report!)
12:40 jonalv I said I think it works
12:41 jonalv it's jsut that I was looking at some code for a project and I don't quite seem to have udnerstood that code
12:41 jonalv lol
12:41 maclean Ah. Misread "no reason to believe" as "reason to believe"
12:41 jonalv oki
12:42 maclean Oh, jonalv : quick question. You worked on molecule db a bit, didn't you?
12:42 maclean (that's not the question)
12:42 jonalv mm yea dot that I want to admit it but yea I did
12:42 maclean The question is : is it feasible to have really, really large databases of just molecules.
12:42 maclean Like billions of them?
12:43 jonalv I am not capcable of answering that
12:43 maclean Actually, sorry that's a stupid question. I guess it's always possible, just not very sensible.
12:43 maclean Just a thought ; nvm.
12:43 jonalv I can tell you that fingerprinting big sets of molecules is not a nice thing to do
12:43 jonalv it's slow
12:44 jonalv :)
12:44 jonalv we wanted fingerprints to be generated during import
12:44 jonalv also, I never had much of a use case sadly enough
12:44 maclean Right. I guess it depends on the fingerprint, though.
12:44 jonalv but the code is around
12:44 maclean Like, some are more expensive to calculate than others.
12:45 jonalv yes of course
12:46 maclean But then, the more useful the fingerprint the more expensive to calculate, probably.
12:46 maclean Anyway, this is not what I'm meant to be doing. Thanks, though :)
12:46 jonalv yea probably
12:47 jonalv ah, you got work todo instead of talking to me?
12:47 maclean Well, I can do both. I just probably shouldn't do a third thing of other work at the same time...
12:48 jonalv makes sense
12:48 jonalv soon™ I will have time to go back and lok at the cdk fingerprints again
12:48 jonalv that\s sort of related
12:51 maclean true - I was surprised by the patch about changing the tanimoto calculation. It seems like two very different ways to compare fps.
12:52 jonalv the count fingerprints_
12:52 jonalv or what was it again_
12:52 jonalv ??
12:52 maclean https://sourceforge.net/tracker/?func=detail&amp​;aid=3313187&group_id=20024&atid=320024
12:53 jonalv yea
12:53 jonalv it\s count fingerpritns
12:53 jonalv I don't think anyone is uisng them
12:53 jonalv I have a branch where I redefine the CDK api to handle count fingerprints better
12:53 maclean I see.
12:53 jonalv sometime I hope to merge
12:53 jonalv although it seems complicated...
12:54 maclean Well, improvements are always welcome.
12:54 jonalv https://github.com/jonalv/cdk
12:54 jonalv thre is a brnach named signFP/rebaserone
12:54 jonalv |-rebaseprone
12:55 jonalv it has both work on the signaturefingerprint and big changes to the fingerprinting api
12:55 jonalv a fingerprint interface for example
12:56 maclean I was confused there for a sec. There's a FingerprintER interface, isn't there?
12:56 jonalv yea
12:56 jonalv but this is for the actual fingerrpint
12:57 jonalv not every fingerprint is good to save in a non sparse BitSet
12:57 maclean Yeah, I see. And the BitFp and CountFp subinterfaces.
12:57 jonalv yea and there are two implementstaion
12:57 jonalv and for bit there are two implmentations
12:57 jonalv the dense based on BitSet and a sparse
12:58 jonalv based on array
12:58 maclean Ah, ok.
12:58 maclean Why would you want to use the sparse one?
12:58 jonalv course my signaturefingerprint is not hashed
12:59 jonalv it\s sparse, but with a big space
12:59 maclean Ok.
12:59 jonalv basilcly it\s a list of integers
13:02 maclean Is there such as thing as a fingerprint of a set of molecules? for comparing sets.
13:02 maclean Sorry, another annoying question, I know.
13:03 jonalv yea I think I saw something similar
13:03 jonalv can\t remember the name of the publication though
13:03 maclean I asked a guy in the lab about this in relation to virtual ligand screening for docking.
13:05 jonalv maclean: I think http://pubs.acs.org/doi/abs/10.1021/ci900159f is in that direction
13:05 maclean tack.
14:44 egonw joined #cdk
16:37 egonw joined #cdk
16:46 maclean joined #cdk
16:55 sneumann_ joined #cdk
17:04 egonw hi sneumann_, maclean
17:07 maclean hi egonw.
17:08 * maclean was just away getting a hot chocolate and a red bull
17:20 egonw np
17:20 egonw doing some online reading
17:20 egonw lounching a bit
17:21 egonw (I want a samsung tab 10.1!)
17:24 maclean Lounching = Slouching + Lounging. Impressive!
17:24 maclean Not an iPad? (or 'fondleslab' as The Register calls it...)
17:37 egonw os/x is for people who like to overpay four a lousy software...
17:37 egonw iPad, sure, if I can install Linux
17:38 maclean heh.
17:39 maclean iOS, anyway. which is even more of a walled garden than OS/X.
17:43 sneumann_ evenin'
17:43 maclean evenin'
18:54 egonw joined #cdk
19:18 sneumann__ joined #cdk
19:28 jbrefort joined #cdk

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