Lars Juhl Jensen
Create an account or sign in to get started
Filter: Blog delicious Flickr Google Reader Last.fm LinkedIn Picasa Web Albums SlideShare Tumblr Twitter Vimeo YouTube FriendFeed
Twitter
posted a message on Twitter
Google Reader
shared an item on Google Reader
Sunday at 12:37 am - Link
very nice analysis Lars - Jean-Claude Bradley
The blog emphasizes that we should always look at absolute risk reduction and not just relative risk reduction, and that number needed to treat (NNT) is important. - Martin Fenner
Almost everyone reporting on these studies has an agenda - I think the government public health services are often worse offenders than companies because they are not required to put a disclaimer that they have a political conflict of interest - Jean-Claude Bradley
Also the subject of Ben Goldacre's column this week: http://www.badscience.net/2008... - Joe Dunckley
Google Reader
shared an item on Google Reader
November 4 at 11:04 am - Link
Google Reader
shared an item on Google Reader
November 10 at 5:42 pm - Link
Ha, coverage of the GCC! Thanx Lars! - Egon Willighagen
"Thanks Michael" would be more accurate - all I did was to press "share" in Google Reader ;) - Lars Juhl Jensen
Thanx Michael! - Egon Willighagen
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
SlideShare
posted a slide show on SlideShare
Network medicine - Integrating drugs, targets, diseases and side-effects
November 6 at 2:30 am - Link
I see what you did here. (Or: I know what the "?" is.) :-) - Michael Kuhn
SlideShare
posted a slide show on SlideShare
Data Integration and Systems Biology
November 6 at 2:39 am - Link
delicious
bookmarked a page on delicious
March 24 at 1:51 am - Link
Database of functional interactions between proteins and chemicals - Lars Juhl Jensen
Yours? - Deepak
Lars, is there a deep-linking syntax? E.g., how does one get "315959" in this direct link to BCR's neighborhood: http://stitch.embl.de/cgi/show.... Custom ID? - Andrew Su
@Andrew: Depends on what kind of ID you have, the kind of stable deep link would be: http://stitch.embl.de/interact... (but this is not a unique identifier of the protein). Guessing BCR_HUMAN as SwissProt id works in this case: http://stitch.embl.de/interact... - Michael Kuhn
"315959" is just an internal id that will change with the versions - Michael Kuhn
Bummer, no other stable external ID then? Ensembl? Refseq? HGNC? for some reason, I find SwissProt to be the hardest to link up to other databases via a unique identifier... - Andrew Su
Ensembl ids aren't stable either, for STRING 7 / STITCH 1 it's Ensembl 36 or so, for STRING 8 it's Ensembl 40. You can try the other ids, for BCR HGNC seems to work: http://stitch.embl.de/interact... . If you want to do it programmatically, there's a "resolve" call in the API: http://string-stitch.blogspot.... - Michael Kuhn
Michael, I'm confused about the Ensembl IDs. This link seems to work fine: http://stitch.embl.de/interact...... If that link syntax is stable relative to whatever version of Ensembl you use, then that will work for me... - Andrew Su
yes, that syntax is stable until a major version upgrade (to STRING 8 / STITCH 2). so if you have Ensembl ids which happen to match ours, you can of course use them - Michael Kuhn
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
Twitter
posted a message on Twitter
Tumblr
posted an entry on Tumblr
October 26 at 11:20 am - Link
Tumblr
posted an entry on Tumblr
October 25 at 4:10 am - Link
Google Reader
shared an item on Google Reader
October 22 at 11:36 pm - Link
Twitter
posted two messages on Twitter
Twitter
posted a message on Twitter
Picasa Web Albums
published a photo on Picasa Web Albums
Second Life: A somewhat happier camper
October 20 at 12:44 pm - Link
Can you spot the two reasons why I am now happier in Second Life? - Lars Juhl Jensen
And I should remember to thank Joanna "Wombat" Scott from NPG for one of them :) - Lars Juhl Jensen
did Joanna give you a tour? - Jean-Claude Bradley
Nope, but she gave me the ability to show in SL that my real name is "Lars Juhl Jensen" :) - Lars Juhl Jensen
Twitter
posted a message on Twitter
Google Reader
shared an item on Google Reader
October 20 at 4:29 am - Link
Neil, Michael, why do you like this paper? I mean, everyone is doing this kind of studies... I mean, it's just like running PCA on your input data... - Egon Willighagen
I like that the post brought the concept to my attention. Barely skimmed the paper :) - Neil Saunders
I just like that Lars has this stuff on the radar. I don't specifically like the paper, as I also think many people do these things. - Michael Kuhn
Hmm, privileged substructures is a pretty old concept. Looks like the paper is applying the concept to new data - Rajarshi Guha
We had a term for it in my old group: "data setje PCA" (and it was typically not positive). - Egon Willighagen
And I just share the paper hoping that Michael Kuhn will read it so that I don't have to ;) - Lars Juhl Jensen
I find it interesting mostly because it sounds new to me. I must be reading all the wrong journals :). Can anyone point to a couple of informative studies related to this ? For example, is this sort of sub-structure analysis combined with predictions of ligand-docking being used to identify drug targets ? - Pedro Beltrao
For target identification, useful papers on the cheminformatics side include http://dx.doi.org/10.1021/ci06... and http://dx.doi.org/10.1080/1062.... Personally, just using docking for target id is not the most reliable approach - Rajarshi Guha
regarding privileged substructures, http://dx.doi.org/10.1021/jm05... is nice. Also useful is http://dx.doi.org/10.1021%2Fci... http://dx.doi.org/10.1021%2Fci... http://dx.doi.org/10.1021%2Fci... and of course the RECAP paper - Rajarshi Guha
Rajarshi, thanks - but sadly, all of the paper you suggest are published by ACS, which means that I don't have access to them :( - Lars Juhl Jensen
One reason I ended a 12 year membership of the ACS earlier this year - Deepak
Yes, unfortunately :( But the papers by Schnuur and Bajorath are very useful, and I haven't seen much in the OA journals - Rajarshi Guha
Thanks again Rajarshi. - Pedro Beltrao
Re the Bayesian papers (also Paolini et al. in NBT): Aren't those high recall rates pretty fishy (>70%). Is this due to bias/redundancy in the libraries that's not corrected for? - Michael Kuhn
I don't think there was (much) bias (though the sizes of the class-wise training sets weren't all equal) in the Novartis paper - one possible reason is that they used nice high quality data. Whether it generalizes to to totally new compounds is always a question. - Rajarshi Guha
Twitter
posted a message on Twitter
Other ways to read this feed:Feed reader