ngn › Comments

Egon Willighagen
#opentox powered toxicity prediction ToxPredict ->
Says "Access forbidden - are you logged in?" - Andrew Lang
Works for me. - Noel O'Boyle
Maybe it is just the Open Tox part that is not working. When I put in a SMILES and then click the "Toxicokinetics >> Gastrointestinal absorption" calculate link, I get the "Access forbidden - are you logged in?" message. - Andrew Lang
Yeah, some computation services need an account... which is, btw, free... - Egon Willighagen
Tomorrow, there is an online tutorial -> - Egon Willighagen
Some resources are intentionally protected, but visible if you have an OpenTox account. Slides / Exercises from the tutorial . - ngn
Any option to remove the 'calculate' option for those services? - Egon Willighagen
Egon Willighagen
Open Standards (Or: "the long list of things that we weren’t required to do") -
Egon says: "important for people to understand: Open Standards (quite like Open Source, and Open Data) are a game changer: they simply collaborating, speeding up development, and changing the way we do things." yepp :-) - Claudia Koltzenburg
Wouldn't that be true for Standards in general? Does it being Open help more? - Rajarshi Guha
I'm tempted to say MDL molfile... it's been really hard for a while to get the ctfile.pdf, though been online freely for a long time now... at the same time we see many interesting, ummm, extensions used in the wild, not following the specs... more importantly, the Open bit reflect to the freedom to implement the specs without strings attached, like patents, royalties, and the sorts... - Egon Willighagen
Yes, I see the issue of royalties etc. But from a practical PoV, an Open Standard will no more prevent variations/extensions than a Closed Standard. (The only thing that prevents that is the "Perfect Standard" :) - Rajarshi Guha
Actually, out of curiosity, are there standards (mainly around life sciences) that require royalties/payment etc to implement & support? - Rajarshi Guha
Good question... I do know of OMIM, an ontology, which is not free: but people are using that as a standard ontology for Open Data (is that even possible? I don't know...) - Egon Willighagen
Regaring OS and variations... I personally agree, which is why I think we need to split off and focus on 'Open Specification' on one side and 'Standards' at the other. To me, the Open Specification is at least as in important, allowing evolution of standards, without it needing expensive, multi-year, consortium-based, ridiculous HTML5-like efforts. I know I disagree here with other BO members, but am fine with that. I for one welcome our forking overlords... also in the Open Standards section of BO. - Egon Willighagen
OMIM as an ontology? - Rajarshi Guha
does that mean any relational table can be an ontology? - Rajarshi Guha
no, I would not say so... not every table has definitions (mostly not, in fact), no hierarchy between concepts, etc - Egon Willighagen
I would say it can; in most cases will translate into a class with properties, foreign keys might be relationships between classes. Hierarchy is not essential (imho), and moreover, there is a way to represent hierarchical structures in relational tables, which could directly translate into hierachy between concepts. At the end these both relational tables ( ) and ontologies are just different representation of domain objects . - ngn
really interesting food for thought, thanks, esp. ngn - Claudia Koltzenburg
ngn... yes, they *can*... but in general are not... - Egon Willighagen
RT @isToxic: #opentox We design new utilities in ToxOtis to consume OPSIN web services ( -
OPSIN based service is running for a while at - ngn
Egon Willighagen
Nina, Barry, how was Ola's talk? Did you like what we managed to do (so far) with #opentox and #bioclipse?
Egon, I liked the demo - it's great to see smooth user interface, being aware how many pieces need to work together on the background :) - ngn
Egon, the demo was not just great, it was one giant step forward on our collaborative capability, a leading indicator of many great things to come. We should organise and explain this story and circulate it in an effective way, so that many audiences understand it. But the reaction from industry users with their needs in mind was very positive :)! And great work all round by you and Ola and OpenTox support in such a short period of a few weeks. Congrats! And you have my vote for my first prize suggestion :) - Barry Hardy
We could make the next demo more spectacular by including toxicity data retrieval and substructures highlighting (#OpenTox API extension needed for the later). - ngn
No code is going to become robust unless someone starts using it in an unintended way ... #OpenTox included ...
True. I think unit tests are the first step here. By forcing to make small, seemingly meaningless tests, you start thinking and using the code in a different way. But in general, the stupider the user, the more robust it becomes :) - Egon Willighagen
PS, you're welcome ;) - Egon Willighagen
As it turned out, unit tests are not enough ... can you imagine what a static field with org.openscience.cdk.ringsearch.AllRingsFinder can do when running server side in parallel threads :) - ngn
Ummm... yeah, we need serious CDK funding... these things must be fixed. For now, different threads, different objects. For both the data model and the helper classes... and if a helper class is statically called, but still uses global (static) variables... that's a very serious, grave bug... please do report if so. I should start blogging about parallelism in the 'Cleaner CDK code' series... - Egon Willighagen
Well, in this case it's ToxTree bug, there was no real reason to have AllRingsFinder static ... - ngn
Egon Willighagen
Workflow Entry: Run OpenTox models on local data
Hey Egon, are you shooting for that prize I promised? Should I bring Pink Floyd discs to Cambridge :)? - Barry Hardy
What prize? Prizes are always welcome :) Well, have a look at the set of workflows I uploaded... Bioclipse can now predict QSAR descriptors from local data (via intermediate data upload) as well as use models to make tox prediction... e.g. Verhaar... - Egon Willighagen
I have a hint for those wanting OpenTox dataset service without a line of code and not using ambit2.war :) I'll not be claiming the prize :) - ngn
Egon Willighagen
Is there an #OpenTox logo we can use for the #friendfeed room?
Yes, but it doesn't look nice fitted in a rectangle - perhaps we could have #friendfeed logo version ? - ngn
OK, uploaded a rectangular one based on what Barry sent me - Egon Willighagen
Antony Williams
RT @egonwillighagen: read the latest comment Fwd @egonwillighagen: chem-bla-ics: pKa prediction, or, how to convert a JCIM paper into Java
looks like it should transfer pretty easily - assuming CDK can handle all the SMARTS - Rajarshi Guha
Rajarshi... that was the problem... we can't... the paper uses unpublished MOE SMARTS... or at least, I could not reproduce their test results :( - Egon Willighagen
The code is available, if you like to play with it... - Egon Willighagen
Or just port Nina's code... - Egon Willighagen
Does Nina's code use the CDK SMARTS code? Or some other implementation? And can she reproduce the paper? - Rajarshi Guha
I forwarded the question to Nina... - Egon Willighagen
No, not the CDK SMARTS code, we have own implementation at There is also maven repository to retrieve compiled jars , e.g - ngn
pKa descriptor code is also available or jar via maven repository or finally via OpenTox REST service - ngn
Related: (I have not read it yet) - Egon Willighagen
@ngn does your code reproduce the results from the Lee paper? - Rajarshi Guha
@rguha you can try the webservice... - Egon Willighagen
yes, I was hoping someone already did the comparison .... lazy friday :) - Rajarshi Guha
The Web Is Dead. Long Live the Internet | Magazine -
"As much as we intellectually appreciate openness, at the end of the day we favor the easiest path. We’ll pay for convenience and reliability" - ngn from Bookmarklet
Other ways to read this feed:Feed readerFacebook