Linked data functional genomics

Linked Data for
functional genomics

Mikel Egaña Aranguren

3205 School of Computer Science
Universidad Politécnica de Madrid (UPM)
28660 Boadilla del Monte
Spain

Ontology Engineering Group (OEG)
http://www.oeg-upm.net

megana@fi.upm.es
http://mikeleganaaranguren.com

http://www.slideshare.net/MikelEganaAranguren/linked-data-functional-genomics

1/12/2011

Index

What is Linked Data?

Publishing Linked Data

Consuming Linked Data

Issues with (Life Sciences) Linked Data

Conclusions

Linked Data for functional genomics


A first step towards the Semantic Web

Query RDF
“Schema” for RDF

“The HTML for data”

Identify things on the net



Linked Data principles

1) Use URIs as names for things

2) Use HTTP URIs so that people can look up those names

3) When someone looks up a URI, provide useful information, using the
standards (RDF, SPARQL)

4) Include links to other URIs so that they can discover more things

http://www.w3.org/DesignIssues/LinkedData.html



With Linked Data we publish data

Semantically: the data model is explicit for computers (RDF triple)

Predicate

Subject Object





Predicate

Subject Object

participates_in

Q5SIF0 ATP binding





Predicate

Subject Object

participates_in

Q5SIF0 ATP binding

Inter-linked: the data is linked to data from other resources over the web

participates_in part_of

Predicate Predicate
Q5SIF0 ATP binding Signal transduction
Subject Subject / Object Object


Global network of Linked Data

part_of

participates_in
ATP binding
Signal transduction
Q5SIF0

is_regulated_by
has_part

Regulation of MAP
Purinergic nucleotide kinase activity
receptor activity



Internet of data rather than documents, a “universal DB”

Find preciselly what we are looking for: direct queries rather than text
processing (SPARQL)

Linking new data is as easy as linking a web page

We can navigate through the data directly (RDF), rather than navigating
through documents that represent data in natural language (HTML)

Build applications that exploit the data

Apply automated reasoning on the data


Linked Open Data (LOD) cloud

http://richard.cyganiak.de/2007/10/lod/lod-datasets_2011-09-19_colored.html



A graph is a collection of RDF triples

A triple store holds different graphs



Human HTML
Content negotiation
Computer RDF



Human HTML
Content negotiation
Computer RDF

Human HTML
Content negotiation
Computer RDF



Biomolecule

rdfs:subClassOf

Vocabulary
Protein Molecular Function
(Ontology)

Instances rdf:type rdf:type

rdf:type
participates_in

Q5SIF0 ATP binding
owl:sameAs

CycB

Dataset
Dataset



Navigate

Query

Meshups



Bio2RDF (http://bio2rdf.org/)

OGOLOD (http://miuras.inf.um.es/~ogo/ogolod.html)

LinkedLifeData (http://linkedlifedata.com/)

HyQue* (http://semanticscience.org/projects/hyque/index.html)

ArrayExpress and Gene expression atlas*
(http://www.ebi.ac.uk/arrayexpress/)

* not really LD, close to LD, but (likely) soon will be full LD



Navigate



Query different resources combining the information

Select all located in Y-chromosome, human genes with known molecular
interactions, which are analysed with 'Transfection'

PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
PREFIX gene: <http://linkedlifedata.com/resource/entrezgene/>
PREFIX core: <http://purl.uniprot.org/core/>
PREFIX biopax2: <http://www.biopax.org/release/biopax-level2.owl#>
PREFIX lifeskim: <http://linkedlifedata.com/resource/lifeskim/>
PREFIX umls: <http://linkedlifedata.com/resource/umls/>
PREFIX pubmed: <http://linkedlifedata.com/resource/pubmed/>

SELECT distinct ?genedescription ?prefLabel
WHERE {
    ?p biopax2:PHYSICAL-ENTITY ?protein .
    ?protein skos:exactMatch ?uniprotaccession .
    ?uniprotaccession core:organism <http://purl.uniprot.org/taxonomy/9606> .
    ?geneid gene:uniprotAccession ?uniprotaccession .
    ?geneid gene:description ?genedescription .
    ?geneid gene:pubmed ?pmid .
    ?geneid gene:chromosome 'Y' .
    ?pmid lifeskim:mentions ?umlsid .
    ?umlsid skos:prefLabel 'Transfection' .
    ?umlsid skos:prefLabel ?prefLabel .
} (http://linkedlifedata.com/sparql)



Query different resources combining the information

We will receive only the triples of that triple store (but we can follow the
links to the triples stored in other triple stores!)



For retrieving triples from other triple stores we need federated
queries:

SERVICE keyword in SPARQL 1.1
PREFIX foaf:   <http://xmlns.com/foaf/0.1/>
SELECT ?name
FROM <http://example.org/myfoaf.rdf>
WHERE
{
  <http://example.org/myfoaf/I> foaf:knows ?person .
  SERVICE <http://people.example.org/sparql> {
    ?person foaf:name ?name . }
}

http://www.w3.org/TR/sparql11-federated-query/



Hypotheses evaluation with HyQue*

http://semanticscience.org/projects/hyque/
PREFIX hybrow: <http://bio2rdf.org/hybrow:>
PREFIX semsci: <http://semanticscience.org/resource/>
PREFIX bio2rdf: <http://bio2rdf.org/ns/bio2rdf:>

select DISTINCT * where {
?event rdfs:label ?label .
?event rdf:type ?event_type .
?event_type rdfs:label ?event_type_label .
?event hybrow:is_negated ?negated .
?event hybrow:physical_context ?event_location .
?event hybrow:physical_operator ?physical_operator .
?event hybrow:agent_a ?actor .
?event hybrow:agent_b ?target .
OPTIONAL {
{ ?actor rdfs:subClassOf ?actor_type } UNION { ?actor rdf:type ?actor_type }
}
OPTIONAL {
{ ?target rdfs:subClassOf ?target_type } UNION { ?target rdf:type ?target_type }
}
?actor semsci:isLocatedIn ?actor_gp_id_location .
?actor_gp_id_location rdf:type ?actor_location_type .
?target semsci:isLocatedIn ?target_gp_id_location .
?target_gp_id_location rdf:type ?target_location_type .
?actor semsci:hasFunction ?actor_gp_id_function .
?actor_gp_id_function rdf:type ?actor_function_type .
}



Meshups: applications that consume LOD

Combining information from different datasets and/or non LOD resources
(e.g. Google maps)

e.g. specific visualisations

e.g. “follow your nose” applications


XML
Flat file

DB
Link
discovery



Announce your data

Comprehensive Knowledge Archive Network (http://ckan.org/)

Semantic Web index (http://sindice.com/)



Why publish our data in the LOD?




It's the links, stupid





Only publish our data, reference the rest: don't need to
duplicate external DB in ours





Only publish our data, reference the rest: don't need to
duplicate external DB in ours

External info is updated independently and we get the benefit
to our dataset because it's linked to it, without extra effort




It's the links, stupid (II)





By using (HTTP) URIs, others can link to us





By using (HTTP) URIs, others can link to us

Increasing the potential for our data to be discovered




It's the semantics, stupid




It's the semantics, stupid

The meaning of our data is easily machine processable due to
RDF (“instances”) and OWL (“schema”)


Issues with (Life Sciences)
Linked Data


Issues with (Life Sciences) Linked Open Data

Provenance (e.g. For Microarray data)

Shared identifiers
http://identifiers.org/
http://sharedname.org

Dataset quality

Ontology modelling

Consensus ontologies

Lack of ontologies

Inference
To generate triples
At query time


Conclusions


Conclusions

Linked Data offers a straight method to publish data
semantically in the current web:

Key 1: use URIs for each and every data item

Key 2: link data items to internal and external data

Key 3: represent data with RDF (and OWL)

Already existing web technology (URI + HTTP) will do the rest
smoothly for us

Knowledge discovery

Knowledge exploitation


Conclusions

Linked Data is here to stay

Already used by many, including governments, BBC, ...

A first usable version of the Semantic Web with great potential

Still issues to be solved in the Life Sciences Linked Data


More information

Semantic Web Health Care and Life Sciences (HCLS) Interest
Group at W3C: http://www.w3.org/blog/hcls

LD Best practices
A. Hogan, A. Harth, A. Passant, S. Decker, and A. Polleres. Weaving
the Pedantic Web. In Linked Data on the Web Workshop (LDOW2010)
at WWW’2010, 2010.

http://patterns.dataincubator.org/book/

Ontology Engineering Group (oeg-upm.net)
Gov. Information on LOD (GeoLinkedData, Aemet, ...)
OGOLOD
Linked Data tools (ODEmapster, ...)


Acknowledgements

I'm funded by the Marie Curie Cofund programme (FP7)

I unashamedly recycled stuff from presentations by Marc-
Alexandre Nolin and Eric Prud’hommeaux

I'm learning a lot at the HCLS IG W3C

NTNU provided the travel/accomodation due to Martin Kuiper's
invitation


Linked data functional genomics

More Related Content

Viewers also liked(20)

Similar to Linked data functional genomics(20)

More from Mikel Egaña Aranguren, Ph.D.(11)

Recently uploaded(20)

Linked data functional genomics