See also: IRC log
%%person1%%: we have several projects making use of linked data at Lilly.
(can someone remind me of the name of the speaker?)
<Susie> William Sanchez is the speaker
scribe: currently we are analysing internal as well as external data
(thanks)
scribe: PubMed and other datasources
... normally this is a very manual process
... we are extending public datasources with proprietary datasources
... we test browsers like Marbles, Visinav
... we test ontologies like UMBEL and VoiD
William: we are doing another project in conjunction with LDD (sp?)
... a document management system
... the product has some in-built metadata capabilities
... there are descriptions for individual datasets, but no inter-dataset relationships
... we want to build a layer on top to integrate
... we use a JDBC interface to access
... we were able to access the datasets with D2R
... the next thing we need to do is to create mapping files. there are thousands of datasets
... another thing we did: we extend spotfire (sp?) to create a SPARQL endpoint
... any user can submit any query to any datasource
... we are in contact with the developers of spotfire to get this into the product
Anja: how are you going to develop the mapping files?
William: if a user finds a specific dataset that he wants to query against, we can respond to that
Susie: what was the file format?
... what you be looking for the group at FU Berlin to make changes to D2R based on your needs?
William: this is quite specific to the issues with JDBC, but we can discuss further
Susie: do you plan to present your work? papers?
William: sure. No particular target at the moment.
Kei: how are the data files related to the databases?
William: there is the possibility to upload the SAS files and make it part of the database
Kei: are analysis results stored backed to the relational database?
William: no, not at the moment
Susie: you recently mentioned scalability limits with one of the UI tools
William: Longwell.
... it puts everything in memory
Susie: how do you link to PubMed?
William: still working on that
Susie: do you think that users would perform a lot of analysis over the linked data, or would it be rather a navigation between datasets?
William: one of the issues is to identify the datasets that a specific project requires.
... after datasets are identified, SAS tools can be used for analysis.
... e.g., "give me all datasets about cardiovascular disease"
<Susie> http://esw.w3.org/topic/HCLSIG/LODD/Questions
Susie: Bosse and I had a call where put together questions. We have 15 in total now. See URL.
... Please have a look at the list and make suggestions
... we should also evaluate what we can already answer with the current LODD datasets, and where we need to add more datasets
... i volunteer on working on the first 3 questions
... with "work on" i mean that we identify which datasets we should add, and how we fare with current data
kei: i would be interested on working on "Are there natural alternatives to this drug? " together with jun
susie: others, please have a look at the wiki and add your name
... we can also make some progress on these questions during the F2F
anja: if someone has "questions about the questions", feel free to ask me about datasets etc.
...: maybe put sub-questions on the wiki
TOPIC --- TCM
kei: we focused on the BioRDF paper recently, so on my side there was not much further progress so far
... matthias has loaded the LODD datasets into the HCLS KB at DERI
... need to think about linkage
... datasets only contained gene symbols, not IDs. not a unique ID.
... we might contact the database curator.
anja: bio2rdf has that online as a dump
jun: between gene symbols and ids?
anja: yes
kei: it is not straight forward, you have to take species into account
susie: if the mapping file is too naive we might explore other sources. william hayes recently spoke about creating a thesaurus for such purposes. not sure if it fits and if they want to share, though.
kei: licensing is an issue.
... all current LODD datasets are open and free to distribute?
susie: we selected datasets based on free licenses, but we need to take care.
<egonw> there is also the problem of license incompatibility
<egonw> http://www.sennoma.net/main/archives/2009/04/an_open_question_about_open_li.php is informative
<egonw> goes into what is and is not allowed in remixing data
susie: some of the datasets had some limitations (e.g. non-commercial)
<egonw> is there an overview of licenses of the current LODD data sets?
<jun> MeSH
jun: mapping to MeSH
anja: we could use the SILK framework
jun: MeSH provide anchors for matching disease IDs
<ericP> SNOMED, mostly
<ericP> there are a bunch of others avail: e.g. UMLS, LOINC
matthias: OMIM only covers a small subset of diseases
susie: we can discuss that at the F2F as well
... there seems to be some overlap between LODD and Pharma Ontology
<ericP> COI is using the Standford Drug Ontology a little as well
<ericP> might want to have helen geeking with y'all
susie: would people that participate remotely be interested in dialing in?
... during the breakouts?
still here, but leaving soon
<ericP> matthias_samwald: if you query the HCLS KB and find probs, please report them
i need to leave now ... eric, could you please add the minutes to the wiki?
bye!
<Susie> http://www.i-semantics.tugraz.at/triplification_challenge
<jun> bye!
<ericP> Susie: next call in four weeks
<ericP> ... haven't heard whether folks want to dial into pharma ontology task
<ericP> AnjaJentzsch: would like to call in, and expect chris to dial in as well
<Susie> http://esw.w3.org/topic/HCLSIG/Meetings/2009-04-30_F2F
<ericP> AnjaJentzsch: given no call in 4 weeks, we should get ontologies which we want to map to
<ericP> ... i could interlink at least drugs, diseases and genes
<Susie> http://esw.w3.org/topic/HCLSIG/PharmaOntology/Roles
<ericP> Susie: pharmaont has creaed a list of q's that folks would ask based on their role
<ericP> Susie: deadline for triplification challenge: 30May