�ݺ�ߣ

Using the Web as the Platform for
Sharing and Consuming
Biomedical Data

Jun Zhao Ph.D.
EPSRC Postdoctoral Fellow
Department of Zoology
University of Oxford

● Use URIs as names for things
● Use HTTP URIs so that people can look up
those names.
● When someone looks up a URI, provide
useful information.
● Include links to other URIs so that they can
discover more things.
Tim Berners-Lee, July 2006
http://www.w3.org/DesignIssues/LinkedData.html

What is Linked Data?
Search for the best public school in
Oxfordshire with tuition fee < ££££ and
find out the houseprice in that
neighborhood

Nutshell of Linke Data (Technical Aspect)

Uniform Resource Identifer
● http://myedubase.db/school001, exEdu:school001
● URI to Linked Data is what URL to Web documents
● Use namespaces to guarantee the uniqueness
● http://myedubase.db/school001
Namespace

● http://edubase.org/school001

Resource Description Framework
● A data model for the Semantic Web
● RDF is a graph-based data model
● Subject, Predicate, Object
● <ox:school001> <ox:is locataed in> <ox:Oxford>
● <ox:Oxford> <ox:is part of> <ox:Oxfordshire>
● <ox:school001> <ox:has national A level ranking> “20”
is located in ox:Oxford is part of
ox:Oxfordshire
ox:Oxford
ox:school001

has national A level ranking
ox:school001 “20”

Now we are on the Web

Everything is identified by a dererenceable URI

Everybody uses the same URI

is located in ox:Oxford is part of
ox:Oxfordshire
ox:Oxford
ox:school001

ox:school001 “20”

Everybody uses the same URI
is located in ox:Oxford
Oxford is part of
ox:Oxfordshire
ox:school001

“20”

Data links
is located in my:Oxford

my:school001

is part of
gov:Oxfordshire
uk:oxford

ox:oxfhigh “20”

Data links
is located in my:Oxford

my:school001 owl:sameAs

is part of
gov:Oxfordshire
uk:oxford

owl:sameAs

ox:oxfhigh “20”

Linking Open Drug Data (LODD)
● A task force of the W3C Health Care Life Science Interest
Group, started since October 2008
● Enrich the Web of Data by publishing drug-related and as
Linked Data
● Investigate the benefits of Linked Data for drug discovery and
biomedical research
● ~ 12 active participants, including researchers and pharmas

Dataset Outgoing links
LinkedCT 220, 569
DrugBank 59, 661
DailyMed 38, 220
RDF-TCM 3, 438
Diseasome 31,065
SIDER 19, 281

Dataset Content Publishing tool Triples
LinkedCT Derived from ClinicalTrials.gov; more than D2R Server 7,036, 000
60,000 trials conducted in the US and other
countries
DrugBank Nearly 5,000 FDA-approved small molecule and D2R Server 767,000
biotech drugs
DailyMed Published by Natonal Library of Medicine D2R Server 164, 300
(NLM); high quality packaging informaton on
4,300 marketed drugs
RDF-TCM 850 herbs, herb-gene and herb-disease Pubby 117, 600
associatons
Diseasome A network of disorders and disorder genes, D2R Server 91, 200
obtained from Online Mendelian Inheritance in
Man (OMIM)
SIDER Informaton on 930 marketed drugs and 1,700 D2R Server 192,500
related side efects
8, 400, 000

Consume linked data
● Datawarehousing
● Convenient, but probamatic for keeping data sync.
● Query federation
● A query mediator to distribute queries to relevant
data sources and then integrate the query results
● Less mature
● Based on SPARQL endpoints rather than the native
Linked Data interfaces

Consuming linked data by following data links

Link traversal-based queries
● Combine link traversal with query execution
● Evaluate part of a query
● Look up URIs returned in the intermediate query
results
● SQUIN.org
● A generic Linked Data query engine
● Can be used as a service: localhost:8080/SQUIN/
● By Olaf Hartig from Humboldt-Universität zu Berlin

Find my school
SELECT ?school ?ranking ?tuition ?price
Where {
<http://my/school1> geo:is_located_in ?loc;
?others geo:is_located_in ?loc ;
uk:national_A_level_ranking ?ranking ;
edu:tuition ?tuition .
?loc house:average_house_price ?price .}

Find schools nearby an area, and their national A level
rankings, annual tuition fees and average house price in
that area.

Find my school
Where {
http
: //...
/s c
h
? ool001

Queried data

Find my school
Where {

Queried data

Find my school
Where {

........
<http://my/school1>
geo:is_located_in
<http://data.gov.uk/location/oxford>
Queried data
........

Find my school
Where {
?loc
http://.../oxford

........
<http://my/school1>
geo:is_located_in
Queried data
........

Find my school
Where {
http
?loc : //...
/ox
? ford
http://.../oxford

Queried data

Find my school
Where {
?loc
http://.../oxford

Queried data

Find my school
Where {
?loc
http://.../oxford
........
<http://my/school100>
geo:is_located_in

Queried data <http://my/school101>
geo:is_located_in
........

Find my school
Where {
?loc ?others
http://.../oxford http://.../school100
http://.../oxford http://.../school101

Queried data

Consume LOD by link traversal
● Pros.
● No need to know all the data sources in advances
● Don't need SPARQL endpoitns for all datasets
● Access the most up-to-date data
● Cons.
● Performance could be a problem
● Requires a URI as the starting point, does not all
kinds of queries

Search for alternative medicines
SELECT DISTINCT ?diseaseLabel ?altMedicineLabel
WHERE {
<http://www4.wiwiss.fu-berlin.de/drugbank/resource/drugs/DB01273>
drugbank:possibleDiseaseTarget ?disease.
?disease owl:sameAs ?sameDisease.
?altMedicine tcm:treatment ?sameDisease.
?altMedicine rdf:type tcm:Medicine.
?sameDisease rdfs:label ?diseaseLabel.
?altMedicine rdfs:label ?altMedicineLabel.
}
What natural alternatives # of query results 7
can be used instead of # of retrieved graphs 28
the drug “Varenicline”? # of accessed servers 6
avg. execution time 0min 46sec

Disea
Drug some
Bank

Linked
Daily CT
med

SIDER RDF-
TCM

The Web of Data
● Allows distributed data publications
● Provides a scalable platform for data sharing
● Creates a web-scale data space
● Offers a new opportunity for data integration and
aggregation

Open issues with LOD
● Best practices for data publication
● Changes of data
● Data synchronization
● “Broken” links
● Data linkings
● Mapping of individuals
● Mapping of concepts
● Provenance and trust
● Large-scale data access and reasoning

Acknowledgement
● W3C HCLS
● Olaf Hartig
●

�ݺ�ߣ

2010 05 edinburgh

More Related Content

2010 05 edinburgh