HCLSIG/LODD/Meetings/2010-01-20 Conference Call

From W3C Wiki
< HCLSIG‎ | LODD‎ | Meetings
Jump to: navigation, search

Conference Details

  • Date of Call: Wednesday January 20, 2010
  • Time of Call: 11:00am Eastern Daylight Time (EDT), 16:00 British Summer Time (BST), 17:00 Central European Time (CET)
  • Dial-In #: +1.617.761.6200 (Cambridge, MA)
  • Dial-In #: + (Nice, France)
  • Dial-In #: +44.117.370.6152 (Bristol, UK)
  • Participant Access Code: 4257 ("HCLS").
  • IRC Channel: irc.w3.org port 6665 channel #HCLS (see W3C IRC page for details, or see Web IRC)
  • Duration: ~1h
  • Convener: Susie


  • Open data follow up - all
  • Data update - Anja, Jun, Matthias, Egon
  • TCM special issue - Matthias, Jun
  • Bio-Ontologies SIG - Susie
  • ACS, AMIA - Egon, Richard
  • AOB


Attendees:EricP, Scott, Egon, Kei, Oktie, Susie, Jun

Apologies: Anja, Bosse

<kei> opendata follow-up

<kei> making data availabe in rdf

<kei> plan to send email to data providers to get their approval and permission

<kei> data update

<kei> egon reported on progress on rdf conversion chembl database at ebi

<egonw> http://pele.farmbio.uu.se/chembl/sparql

<kei> links of drugs to targets and clinicaltrials

<kei> not only drugs on market but also failed drugs

<egonw> http://chembl.blogspot.com/

<kei> who contributed some of the data to chembl?

<kei> susie: how to use sparql to enable meaningful searches (e.g., substructure)

<kei> egon: find small molecules (drugs) for particular targets

<ericP> http://pele.farmbio.uu.se/chembl/sparql endpoint seems to need a tickle

<kei> egon: use sparql to do substructure mining/subsubstructure comparison

<ericP> [[ select distinct ?Concept where {[] a ?Concept} limit 2 ]] taking forever and it used to be instant

<kei> susie: this might be an interesting area to explore the feasibility of substructure search

<kei> susie: link chembl to other resources such as TCM and drugbank

<kei> egon: yes, this is important

<kei> scott: question of substructure search: different approaches such as chemblast by TN Bhat in the context HIV virus

<kei> scott: characterize different types of structures using rdf

<mscottm> http://xpdb.nist.gov/pdb_chem_blast/help.html

<kei> egon: they tried to fingerprint molecules using chemblast

<kei> scott: start with basic structures and refine them in the hierarchy

<mscottm> http://bioinfo.nist.gov/SemanticWeb_pr2d/chemblast.do

<kei> egon: good start, but finger printing generates noise

<kei> egon: the noise problem needs to be addressed

<kei> susie: oktie has updated linkct dataset

<egonw> ericP: yeah, I think I might have messed up the index :)

<egonw> ericP: I need to check how many triples I have right now...

<kei> oktie: updated based on new data extracted from CT.gov

<egonw> but in RDF/XML format... it's about 2GB of data

<kei> oktie: interventions have erroneous links to trial data

<kei> oktie: CT has fixed the problem so the data quality is better

<ericP> egon, (checkign my comprehension) i understood from what you said that more than one substructure leads to the same fingerprints; fingerprints are a first pass at restricting to substructures. correct?

<egonw> correct

<OktieH> http://queens.db.toronto.edu/~oktie/linkedct/

<kei> oktie: latest dump available

<kei> oktie: should update the HCLS KB endpoint at DERI

<egonw> ChEMBL-RDF example: http://pele.farmbio.uu.se/chembl/snapshot.php

<kei> oktie: after the update, many things have changed so old links may not work

<egonw> source code: http://github.com/egonw/chembl.rdf

<egonw> issues, comments, etc can be filed there too

<kei> oktie: Anja should look at her data that has links to CT

  • mscottm doesn't think that the chemblast approach produces the fingerprints that Egon addressed but characterizes substructures precisely. Main problem seems to be that it is difficult to do exhaustively.

<egonw> mscottm: I will read the paper and comment on it next meeting

<kei> susie: Anja sent her apologies for missing today

<kei> susie: asked oktie to contact Anja

<kei> oktie: links between drugs and interventions in CT

<kei> oktie: types of links should not be owl:sameAs because intervention is not just a drug (but drug dosage sometimes)

<kei> oktie: this link semantics problem should be addressed as soon as possible

<kei> susie: similarity or dissimilarity between drug and intervention

<kei> susie: would it help to use a more standard terminology/vocabulary?

<kei> susie: adoption of consisting naming in links

<kei> oktie: create another entity for drug (e.g., generated from interventions)

<kei> oktie: this might ease maintenance/management of links

<kei> susie: link to the standard definition of drug (in some ontology)

<kei> oktie: it's helpful

<kei> susie: tcm special issue

<SusieS> kei: deadline for tcm is the end of the month

<SusieS> kei: deadline may be extended, but very uncertain at the moment

<SusieS> kie: have been in touch with matthias

<SusieS> Kei: matthias has been talking to Huajun

<SusieS> kei: but don't have a recent update

<SusieS> kei: Jun is also working on the paper

<SusieS> kei: but also haven't heard from her

<SusieS> kei: the special issue has long research paper and shorter commentary paper

<SusieS> kei: will mention to Matthias

<egonw> drug terms in the ChEBI ontology:

<egonw> http://www.ebi.ac.uk/chebi/searchId.do?chebiId=CHEBI:23888

<kei> egon: drug roles are described in chebi

<kei> susie: The drug concept was discussed in the TMO effort

<kei> susie: TMO could potentially be used for sharing information about drugs

<kei> susie: network of collaborations is key to future companys' success

<kei> susie: ontologies are important in the collaborative context

<kei> susie: information captured chebi might be of interest

<kei> susie: there is a drug ontology in stanford

<kei> scott: saw Samson Tu who was involved in drug ontology

<kei> scott: Samson seemed to be willing to work with us

<kei> scott: a follow up might work

<kei> susie: Trish might also be able to follow up with Samson

<kei> susie: chebi focuses on small molecules at the moment, but we might need a broader context

<kei> susie: larger molecule, RNAi, ....

<kei> oktie: chembl and chebi are linked ...

<kei> susie: deadline for bio-ontology sig is approaching

<kei> susie: people have been busy

<kei> susie: upcoming meetings?

<kei> ACS meeting -- egon might be able to give update

<kei> end of march

<kei> susie: identify additional data sources for linking and mapping

<kei> susie: how people use and interact with the converted data is important

<kei> susie: visualization of data is also important

<kei> susie: can data providers convert their data in rdf or linked data?

<kei> susie: technologies are maturing/improving

<kei> susie: governments are more involved in the data sharing process

<kei> susie: invite representatives from government to give talks.

<kei> scott: uniprot available in rdf

<kei> scott: these are linked data

<kei> kei: data size and performance

<mscottm> kei: Next Generation Sequencing, etc. provide a challenge when trying to represent data in RDF

<kei> susie: data sources are very diverse and mappings are very complex

<kei> susie: experimental/clinical data contexts need to be captured in some way

<kei> susie: levels of granularities

<kei> susie: reasoning/inferencing using triplestores

<kei> eric: pre-processed inferencing

  • ** zhaoj [chatzilla@] has joined #hcls

<kei> susie: linked data framework use cases -- sparql queries

<egonw> http://saml.rilspace.com/content/initial-performance-comparison-pellet-vs-prolog-in-bioclipse

<egonw> my student Samuel very much likes to here about time consuming queries and compare SPARQL and prolog there

<kei> susie: how to explore and aggregate information (e.g., entity-based)

<kei> susie: interesting issues to explore: inferencing, data mapping, user-interface paradigm

<kei> eric: warehouses are still prevanlent, distributed/federated queries

<kei> scott: how to make use of the data in a meaningful way ...

<kei> susie: another area: how far the semantic mapping can go?

<kei> susie: federated queries: how they can be done?

<kei> eric: when semantics are not working, string matching is a backup approach ...

<kei> egon: reasoning compare sparql and prolog

<egonw> http://saml.rilspace.com/

<kei> scott: prolog can be a nice underpining to reasoning

<kei> scott: reasoning and rule integration is complicated

<kei> eric: new reasoning features may be added to sparql

<kei> scott: use of property chaining in sparql

<kei> susie: need to wrap up

<kei> susie: how to come up with best practices in terms of reasoning ...

<kei> susie: we can start with experimental data ...