Camelia, the Perl 6 bug

IRC log for #bioclipse, 2009-11-09

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

All times shown according to UTC.

Time Nick Message
00:38 egonw_ joined #bioclipse
06:55 egonw joined #bioclipse
08:00 sneumann_ joined #bioclipse
08:12 Gpox joined #bioclipse
08:15 mgerlich joined #bioclipse
08:45 jonalv joined #bioclipse
08:48 jonalv oh hai
08:48 zarah hello jonalv, you fantastic person you
08:55 egonw joined #bioclipse
08:55 sneumann joined #bioclipse
09:14 edrin joined #bioclipse
09:15 sneumann hi there, minor suggestion:
09:16 sneumann what about adding release dates to the files on http://pele.farmbio.uu.se/bioclipse-devel/
09:16 zarah sneumann's link is also http://tinyurl.com/l7lteh
09:16 sneumann and the news items on http://www.bioclipse.net/
09:31 masak joined #bioclipse
09:40 egonw yes, agreed
09:40 egonw please file them at bugs.bioclipse.net
09:41 egonw olas needs to do this, and he is not yet online
09:47 sneumann Bug 1721
09:57 egonw @pz 1721
09:57 zarah http://pele.farmbio.uu.se/cgi-bi​n/bugzilla/show_bug.cgi?id=1721
09:57 egonw tack
10:07 edrin http://labs.knime.org/webservice-node
10:07 zarah edrin's link is also http://tinyurl.com/yenk8zw
10:07 edrin interesting, the web service plugin for knime was generated by AstraZeneca
10:11 egonw it says *sponsored*
10:11 egonw I think AZ payed for the development
10:20 edrin oh ok
10:57 samuell joined #bioclipse
11:15 * jonalv @ lunch
12:32 CIA-51 bioclipse.blipkit: Samuel Lampa master * r7fb87c7 / plugins/net.bioclipse.blipkit/src/net/biocl​ipse/blipkit/business/JPLQueryWrapper.java : Improved checking of atom term type - http://bit.ly/15pYAq
13:03 * samuell is grabbing coffee
17:17 egonw joined #bioclipse
18:06 sneumann joined #bioclipse
18:33 sneumann_ joined #bioclipse
18:49 egonw joined #bioclipse
18:55 sneumann_ joined #bioclipse
19:10 egonw samuell: ping
19:10 egonw samuell: are you still in the office? if so... I have a question :)
19:10 samuell egonw: pong
19:12 samuell Yes?
19:12 egonw I have nailed an invoice on my wall
19:13 egonw two papers on top of the schema of jonalv's db
19:13 egonw the top one is of UPS
19:13 egonw the second one of Ingram Transportation or so...
19:13 egonw can you look at that second paper, and look up the account number please?
19:16 sneumann_ joined #bioclipse
19:17 samuell egonw: Yep, sure!
19:19 samuell egonw: The top one has a "INGRAM" logo, but sais nothing about UPS, which the bottom one does
19:20 egonw ok, than the reordered them
19:20 egonw the one with the ingram logo
19:21 egonw does that give an account number
19:21 samuell Ok... trying to find "Accound number"
19:22 samuell Can find waybill no and shipmentID, but no account number, it seems
19:22 samuell Hmm... PO Number
19:22 samuell Don't know what that means...
19:22 egonw ok, can you scan them and email the PDFs?
19:22 egonw will send that to the US office the
19:22 egonw then they can figure out what the find useful
19:23 egonw should have scanned myself... should have though of that
19:23 egonw have you done scanning before?
19:23 samuell No... It's done with the "kopiator" I guess?
19:28 egonw great! received! many thanx!
19:30 samuell np!
19:31 egonw samuell: you can pose your question about using reasoning for experimental scientific numerical data on semanticoverflow
19:31 egonw interesting to see what they think of it :)
19:31 samuell yeah well, but was it really a question?
19:31 egonw I hope so...
19:31 samuell Or how would I formulate it as a question?
19:31 egonw I have not seen any facts of experiment results
19:32 samuell I think I was rather putting forth some arguments.... though I might indeed explain better what I mean, and give references
19:32 samuell (I think I do actually have references for that... )
19:33 sneumann_ joined #bioclipse
19:34 samuell I guess, when trying to figure out the capabilities of a certain technology, you can't rely on pure experiments... you have to grasp how it works, and have an idea...
19:34 samuell Of course, then you have to test it out
19:34 samuell to see that you'r not thinking crap.
19:34 samuell Don't you think?
19:36 samuell With NMR data, I can't even come up with an idea about how to do a typical reasoning problem...
19:37 samuell The data would need preparation with number-crunching technoloiges (statistics based classifiers)
19:37 samuell which could turn it into more typical semantic data
19:38 samuell ...for example saying that this and this subspectrum is actually of a type that points to this and this kind of molecule et.c.
19:40 samuell I think it might actually be almost as far from a suitable semantic representation as you can possibly represent a molecule
19:41 samuell ...while chemical graphs seems to be the opposite
19:41 samuell That is... very close to the kind of representations that semantic technologies are aimed to work with.
19:41 samuell The difference between these two representations is interesting though
19:42 samuell But....
19:43 samuell ... the transition from semantic-void number data (NMR Shift) to semantic ditto (Chemical graphs) I would say is arguably best done with existing number-crunching technologies, simply because the raw data consists of numbers but not semantics.
19:46 egonw is prolog not able to make a statement about the numerical distance between no numbers?
19:46 egonw and make that true or false?
19:46 samuell egonw: Yep, it is...
19:46 egonw proximate(x,y) := abs(x-y) < epsilon
19:47 samuell So... in principle it is indeed possible to do (at least some simple) number crunching with prolog...
19:48 samuell but I'm thinking that the kind of number crunching methods required in order to transfer the number data (NMR Shifts) into really semantic data (for example a a chemical graph), is a whole field of study... and probably would take up a whole degree project
19:49 samuell Isn't it so?
19:49 egonw why is are the current chemical shifts not semantic data?
19:49 egonw btw, check out SCOVO
19:50 samuell Ok!
19:50 samuell Well, it is indeed represented in a semantic format yes... but...
19:51 egonw ok, now it gets interesting...
19:51 egonw please continue
19:51 egonw and I think you may have your introduction
19:51 samuell Think about a researcher, without any computational tools ... just with some nmr shift data
19:52 egonw here I am
19:52 samuell Yep :)
19:52 samuell He would not EASILY be able to say that this is a such molecule, and this is a such...
19:52 samuell But
19:53 egonw wait
19:53 samuell If he can TRANSFER the information in the NMR Shifts into a more ACCESSIBLE form, that more closely relates to HOW he REASONS himself
19:53 egonw I don't understand your statement
19:53 egonw [11/09/09 20:52] <samuell> He would not EASILY be able to say that this is a such molecule, and this is a such...
19:53 samuell e.g. by (through painful calculations) drawing the chemical graphs of the molecules, and saving thenm
19:53 egonw what can't he easily do ?
19:54 samuell egonw: Ok... well, I understand how you mean...
19:54 samuell I mean... he could say that if he was an NMR expert
19:55 samuell But what if he knew chemistry only in terms of chemical graphs? (Which one could say is a more general representation, no?)
19:55 egonw why would he know chemistry only in terms of graphs?
19:56 egonw why would he not know that 13C shift of around 170 is a carbonyl ?
19:56 egonw 145 another C.sp2...
19:56 samuell Hmm...
19:56 egonw 50 C attached to a C-O ...
19:57 egonw is your point that you want that nmrshiftdb data too?
19:57 samuell egonw: Yeah, I would indeed need those classification rules
19:57 samuell Well, I think you are right...
19:57 samuell and I am wrong....
19:58 samuell But... as said, I would need access to that kind of classification rules
19:58 samuell in order to do anything useful.... AFAIS
20:00 samuell Hmm... good example....
20:00 samuell I had obviously forgotten some things we were talking about
20:01 samuell (That is, how one would be able to connect typical  features of a spectrum to certain functional groups et.c.... we were talking about that before, right?)
20:02 samuell But, how can I find such data? (How to classify molecules according to spectra?)
20:03 egonw reference knowledge
20:03 samuell Ok
20:03 egonw mom
20:04 egonw http://www.cem.msu.edu/~reusch/OrgPage/nmr.htm
20:04 zarah egonw's link is also http://tinyurl.com/y9g6xjy
20:04 samuell Thx!
20:05 samuell That would indeed be a good reasoning example
20:05 samuell ...to implement those ranges as prolog classification rules.
20:08 samuell So... my previous argumentation was not valid in this case.... since obviously people have already done enough statistics and number-crunching in order to derive such classification rules :)
20:08 egonw well, science is about reproducing...
20:09 egonw there would be nothing wrong with number crunching in a new way...
20:09 egonw reasoning may not sounds the most logical things...
20:09 egonw certainly not if the reasoning will not beyond calculating a mean...
20:09 egonw but when the reasoning is comining things...
20:10 egonw like #C query == #C db match + proximate(shift1_query, shift1_dbmatch) + ....
20:10 egonw that way you can combine things: numerical and categorical
20:11 egonw there are not many method that can do that...
20:12 samuell True... making knowledge interoperable.
20:13 samuell Yeah, well. I think I see your point.
20:17 samuell So, it's not only about doing semantically advanced reasoning, but about making use of the data we actually have... in whatever format it might be, and being able to combine it with other data... which might require some shift of representation et.c.
20:19 samuell Did I understand you right?
20:20 samuell Anyway... thanks for the discussion, and your patience :)
20:33 egonw yes...
20:34 egonw also think of the data you were originally considering to use
20:34 egonw the assay data from mats...
20:34 egonw that's even far more numerical than the nmrshiftdb data
20:34 samuell Hmm...
20:35 samuell Yeah, that's probably true.
20:36 samuell We had some thoughts though, to combine it with clinical data
20:36 samuell that is... we could connect the "correct answer" to specific characteristics of the numerical data
20:37 samuell But still, it's true that it would involve probably a large amount of number crunching
20:37 egonw yes, so, you can now define some subquestions of your original 'performance' questions...
20:39 samuell Yes, probably... I will need to think about the formulation of those questions
20:54 samuell Hmm... On a different matter, your conclusion about combining things being one of the points makes me think a bit.
20:56 samuell Since it is much a problem of technical barriers also, prolog seems less and less suitable as the overarching integrator. At least per se. Bioclipse would be the more natural place to do the actual integration, and now, with blipkit it is getting a bit confusing about where to let the core integration / reasoning happen.
20:56 samuell Another problem is that to much investment in prolog as a semantic technology is detrimental to the SemWeb effort, since it is actually promotes the building of a separate technology stack, from the main one, which has SWRL as it's rule language for the same goal [I got a ref for this].
20:56 samuell Prolog is a mature technology though.... so I'm getting struck by the thought that Bioclipse should use OWL / SWRL as it's internal knowledge integration glue (SWRL would need to be added to be able to  capture things like circular structures), and then only possibly execute the SWRL in prolog (which might be possible)
20:57 samuell (BUT ... I will indeed finish my project integrating prolog anyway, no worry :) )
20:58 egonw no worries
20:58 egonw it's merely a 20wk project
20:58 egonw you do not need to solve the world problems...
20:58 egonw and, yes, I agree... I think Bioclipse is an excellent platform for semantic computation
20:58 egonw but Bioclipse is very much about bridging worlds...
20:59 egonw and bridging functionality and services...
20:59 egonw that's where Bioclipse-Blipkit comes in...
20:59 egonw to make it scientifically interesting, we need to experimentally determine when to stay on the Bioclipse side, and when to move into the blipkit world...
21:00 egonw I am hoping your project will shed some light on that
21:00 samuell Yep, exactly
21:01 * egonw is going offline...
21:02 egonw tired
21:02 egonw slept only 5 hours last night
21:02 samuell It's getting late
21:03 samuell Yeah, 5h is no good amount
21:19 samuell Do yo know BioGateway well? ( http://www.semantic-systems-biology.org/biogateway )
21:19 zarah samuell's link is also http://tinyurl.com/ya7g2rl
21:20 samuell (Just wondering, as I came across it... don't remember if you've mentioned it earler)
21:29 samuell Ok... gotta go in a minute

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