Bio ontologies and semantic technologies

Introduction to Bio Ontologiesand The Semantic Web

M. DevisscherBiological Databases

Overview

• Bio ontologies• Semantic technologies

• Practical sessions: – Protégé and a bio database– DYI SPARQL endpoint

Introduction

• Ontologies: what are ontologies ?

• Ontologies in the bio domain: OBO Foundry• Ontologies in the semantic web

• OBO• RDF, IRI, TTL, SPARQL, OWL

What is an ontology ?

• Ontology = a specification of a conceptualization (Gruber 1993)

• In practice: controlled vocabularies– Disambiguation (e.g. Bank, Running)– Language/species independence

• Very useful in biology – complex hierarchies of terms

Ontologies in the bio Domain

• OBO Foundry -‐ open Biological andBiomedical Ontologies

• Common principles• List of ontologies at http://www.obofoundry.org

• OBO is also a data format .obo

SideTrack – The Gene Ontology

• The mother of bio-‐ontologies: the GO– Oldest bio – ontology– Many practical applications:• Cross species studies• Term abundance studies

• GO is an OBO ontology


• Collection of terms


• Relationships between terms:– Subsumption: is_a– Partonomic: part_of

• These terms are transitive• Terms form a DAG (directed, acyclic graph)• Some information can be inferred




• Knowmore: www.geneontology.org• AMIGO : the GO browser

Gene Ontology Annotation

• Gene ontology annotations GOA = entities labeled with GO terms– E.g. Uniprot-‐GOA

Semantic Technologies

• The semantic web: Tim Berners Lee et al, Scientific American 2001


• W3C: a set of specificationshttp://www.w3.org/standards/semanticweb/

• A mature toolset– Dedicated data formats– Storage– Query language


• Basic data element = a Triple– A mini sentence– Contains three Terms:• Subject Predicate Object


• Representation of triples– Basic data format: RDF/XML– All data expressed in RDF (Resource DescriptionFramework)

– Several compatible syntaxes: TTL (Terse Triple Language) most human readable

Example

The Turtle Syntax

• Basic Triple

<http://bioinformatics.be/entities#martijn><http://bioinformatics.be/relations#has_favorite_beer><http://bioinformatics.be/entities#karmeliet>.

The Turtle Syntax

• Prefix

@prefix b4x: <http:bioinformatics.be/terms#>b4x:martijn b4x:has_favorite_beer b4x:karmeliet.

The Turtle Syntax

• Predicate lists

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .b4x:martijn b4x:has_favorite_beer b4x:karmeliet;

foaf:name “Martijn Devisscher”.

The Turtle Syntax

• Object lists

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .b4x:martijn b4x:has_favorite_beer b4x:karmeliet,

b4x:chimay_blauw;foaf:name “Martijn Devisscher”.

IRI’s and Literals

• Terms can be either IRI’s, Literals or blank nodes• IRI = Internationalized Resource Identifier• Unique id – a virtual URI– Example: http://bioinformatics.be/terms#martijn– There is no requirement for resolving– Now: Open Data initiatives: please do use resolvableURI’s http://linkeddata.org

– Unique identifiers can be registered on http://identifiers.org

Introduction

• Literals: can be typed, allowed types from the XSD namespace:– E.g. “This is a string example”^^xsd:string– E.g. “5”^^xsd:integer

• IRI’s are used for entities and attributes• Literals are used for attribute values thataren’t entities

The Turtle Syntax

• Typed literals

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .b4x:martijn b4x:has_favorite_beer b4x:karmeliet,

b4x:chimay_blauw;b4x:length “184”^^xsd:integer;foaf:name “Martijn Devisscher”^^xsd:string.

The Turtle Syntax

• Blank nodes

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .b4x:martijn b4x:has_favorite_beer b4x:karmeliet,

b4x:chimay_blauw;b4x:length “184”^^xsd:integer;foaf:name “Martijn Devisscher”^^xsd:string;b4x:owns_cat [ b4x:color “Gray” ].

Classes and Individuals

• rdf:type

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .b4x:martijn rdf:type foaf:Person.

Classes and Individuals

• Shorthand: a

@prefix b4x: <http:bioinformatics.be/terms#> .@prefix foaf: <http://xmlns.com/foaf/0.1/> .b4x:martijn a foaf:Person;

foaf:knows b4x:geert.b4x:geert a foaf:Person.

Example

<http://xmpl/entities#martijn><http://xmpl/relations#has_favorite_beer><http://xmpl/entities#karmeliet>.


• Sets of triples form a Graph

Graphs

• Triples are building blocks of Graphs

• Combining sets of triples allows the construction of arbitrarily complex graphs

b4x:martijn b4x:karmeliethas_favorite_beer

Add meaning !

• Reuse terms from existing, well definedvocabularies – ontologies (foaf, dc, go, so)

• Describe new terms = Ontologies

• Contain– A crisp human definition– Some machine readable facts

Metadata

• Ontologies are also described in RDF– RDFS: RDF -‐ Schema– OWL: Web Ontology Language– Also expressed in RDF

• For clarity, file extension can be .rdfs or .owl

RDFS Essentials

• Descriptions– rdfs:label– rdfs:comment

RDFS

• Relationships between properties, classes– rdfs:Class– rdfs:subClassOf– rdf:Property– rdfs:subPropertyOf– rdfs:range– rdfs:domain

RDFS: Example

@prefix rdfs: <http://www.w3.org/2000/01/rdf-‐schema#>.@prefix foaf: <http://xmlns.com/foaf/0.1/> .@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .b4x:karmeliet a b4x:Trappist .b4x:Beer a rdfs:Class .b4x:Trappist a rdfs:Class .b4x:Trappist rdfs:subClassOf b4x:Beer .b4x:has_favorite_beer a rdf:Property ;

rdfs:domain foaf:Person ;rdfs:range b4x:Beer .

b4x:Beer rdfs:subClassOf b4x:Drink .

Analogy

• RDF = database = data• RDFS/OWL = schema = metadata

• Both are described in RDF, but have a different scope


• Inference– Enhance dataset using knowledge frommetadata(e.g. rdfs, owl)

• Types of inference engines– RDFS inference• RDFS entailment regime

– OWL inference• Under active research• Engines exist for specific subsets of OWL (OWL-‐DL)

RDFS Entailment

RDFS: Inference

b4x:kevin b4x:has_favorite_beer b4x:stella

Q: What can we infer from this using RDFS entailment ?

RDFS: Inference

b4x:kevin b4x:has_favorite_beer b4x:stellaInferred triples:b4x:kevin a foaf:Person [from domain]b4x:stella a b4x:Beer [from range]b4x:stella a b4x:Drink [from subClassOf]

DuckTyping

• Watch out with inference !

Example: You want to express that people canhave lengths

b4x:length a rdf:Property;rdfs:domain foaf:Person;rdfs:range xsd:integer.

DuckTyping

• Problem:

ex:VW_Transporter b4x:length “600”^xsd:integer.

• Would infer that VW_Transporter is a Person !• This is called DuckTyping

If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck

Task

• Find a solution: express in rdfs that people can have lengths

Task

• Find a solution: express in rdfs that people can have lengths

b4x:havingLenght a rdfs:Class.b4x:length a rdf:Property;

rdfs:domain b4x:havingLength;rdfs:range xsd:integer.

foaf:Person rdfs:subClassOf b4x:havingLength.

Storing RDF

• As an RDF file for download• In a Triplestore– Database optimised for storing triples– Examples: BlazeGraph, Fuseki, Sesame


• Querying over RDF data: SPARQL• Cool features:– Distributed querying = actual distribution of data and computing resources

– SPARQL/Update: modify data

• SPARQL endpoints: SPARQL over HTTP

SPARQL Query Syntax

• First example:

SELECT ?subject ?predicate ?object WHERE {?subject ?predicate ?object.

}

(Generally not a good idea as it will pull down the whole dataset)

Binding variables

Graph matching

?

SELECT ?person WHERE {?person b4x:has_favorite_beer b4x:karmeliet

}

?

SPARQL Query Syntax

• Limit result size :

SELECT ?subject ?predicate ?object WHERE {?subject ?predicate ?object.

} LIMIT 10

SPARQL Query Syntax

• Find all classes:

PREFIX rdfs: <http://www.w3.org/2000/01/rdf-‐schema#>SELECT ?class ?label WHERE {

?class a rdfs:Class.?class rdfs:label ?label.

}

(This will only retrieve classes that have a label)

SPARQL Query Syntax

• Find all classes:


?class a rdfs:Class.OPTIONAL {

?class rdfs:label ?label.}

}

SPARQL Query Syntax

• Find all classes that contain “duck” in the label:


?class a rdfs:Class.?class rdfs:label ?label.FILTER( CONTAINS (str(?label) , “duck” ) )

}

SPARQL Query Syntax

• Make it case insensitive:


?class a rdfs:Class.?class rdfs:label ?label.FILTER( CONTAINS ( UCASE(str(?label)) , “DUCK” ) )

}

SPARQL Query Syntax

• Search in specific graph:

PREFIX rdfs: <http://www.w3.org/2000/01/rdf-‐schema#>SELECT ?class ?label FROM <http://example.org/animals>WHERE {

?class a rdfs:Class.?class rdfs:label ?label.FILTER( CONTAINS ( UCASE(str(?label)) , “DUCK” ) )

}

SPARQL Query Syntax

• Search in specific graph:


GRAPH <http://example.org/animals> {?class a rdfs:Class.?class rdfs:label ?label.FILTER( CONTAINS ( UCASE(str(?label)) , “DUCK” ) )

}}

SPARQL Query Syntax

• Can also search for graphs :

PREFIX rdfs: <http://www.w3.org/2000/01/rdf-‐schema#>SELECT ?g WHERE {

GRAPH ?g {?class a rdfs:Class.?class rdfs:label ?label.FILTER( CONTAINS ( UCASE(str(?label)) , “DUCK” ) )

}}

Summary: Querying RDF data

RDF Data InferenceEngine

RDFS/OWL

RDF Data

Inferred

SPARQLEndpoint

• Basic data element = a Triple– A mini sentence– Contains three Terms:– Subject Predicate Object

• Example:

<http://xmpl/entities#martijn><http://xmpl/relations#has_favorite_beer><http://xmpl/entities#karmeliet>.

Take home Summary

• Combine triples to represent knowledge

• Use terms from ONTOLOGIES

– COMMON VOCABULARIES– POSSIBLE TO INFER

MEANING• OMIABIS• OBIB• SNOMED/ICD• MESH

?

• SPARQL searches for patterns

?

Interoperability between OBO andSemantic Technologies

• Originated from two separate academic worlds• Computing applications of OBO mainlyconsistency checkingand overrepresentationanalysis

• Semantic Technologies: much broader toolset

• Interoperability ?– Direct offering in both formats– Automatedmapping

Where to find ontologies

• OBO Foundry• Bioportal; NCBO• Biogateway• Bio2RDF

Where to find RDF data

• Google for SPARQL endpoint• => e.g. EBI databases

• Non biological: DBpedia

How about Tim Berners Lee’s vision

• We’re not there yet, but for bio data we’regetting quite close– The explicitome– Crowd sourcing– Nanopublications

SPARQL in PRACTICE

SPARQL : Recap

PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>SELECT ?label FROM <http://graphName> WHERE {

?x rdfs:label ?label.FILTER ( CONTAINS(?label, “dimethylalinine”) )

} LIMIT 10 ORDER BY ?label

SPARQL : Recap




• FIND the pattern ?x rdfs:label ?label.

SPARQL : Recap





• BIND variables ?label, ?x

SPARQL : Recap





• BIND variables ?label, ?x• RETRIEVE variable ?label

SPARQL : Recap





• BIND variables ?label, ?x• RETRIEVE variable ?label• PREFIX: replace rdfs:label by <http://www.w3.org/2000/01/rdf-schema#>

SPARQL : Recap





• BIND variables ?label, ?x• RETRIEVE variable ?label• PREFIX: replace rdfs:label by <http://www.w3.org/2000/01/rdf-schema#>• FILTER results to labels containing “dimethylalinine”

SPARQL : Recap





• BIND variables ?label, ?x• RETRIEVE variable ?label• PREFIX: replace rdfs:label by <http://www.w3.org/2000/01/rdf-schema#>• FILTER results to labels containing “dimethylalinine”• LIMIT results to first 10 matches ordered by label

SPARQL : Recap

DESCRIBE <http://rdf.wikipathways.org/Pathway/WP1425_r74390/WP/Interaction/e077e>

• Useful short query to get direct links from/to a given node

SPARQL REFERENCE

http://www.w3.org/TR/sparql11-‐overview/

Running SPARQL• From a web interface

• From a web interface• Using http

– HTTP GET

– HTTP POST : for larger query strings– Headers determine response type (JSON, XML, HTML)

http://…/sparql?default-graph-uri=<http://graphName>&query=URLENCODEDQUERYSTRING

Running SPARQL

BIO-‐ONTOLOGIES

BioPortal

Access

• From the web interface !• SPARQL endpoint: using API key; on request • Running a local copy: download VM image; on request

Exercises

• Find a term• Find ontologies containing a term• Browse some ontologies• Check the NCBO annotator !

BIO-‐DATA

EBI RDF Resources

EBI RDF Resources

Ensembl

Exercise

• From uniprot find proteins that are annotated with a given Gene Ontology term

PREFIX up:<http://purl.uniprot.org/core/> PREFIX taxon:<http://purl.uniprot.org/taxonomy/> PREFIX rdfs:<http://www.w3.org/2000/01/rdf-schema#>PREFIX obo:<http://purl.obolibrary.org/obo/>SELECT * WHERE {

?protein up:classifiedWith obo:GO_0004499.?protein up:organism taxon:9606.

}

http://sparql.uniprot.org

Exercise

• From Expression Atlas find proteins that are differentially expressed (P < 1e-‐12) in Crohn’sdisease

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>PREFIX owl: <http://www.w3.org/2002/07/owl#>PREFIX dcterms: <http://purl.org/dc/terms/>PREFIX obo: <http://purl.obolibrary.org/obo/>PREFIX sio: <http://semanticscience.org/resource/>PREFIX efo: <http://www.ebi.ac.uk/efo/>PREFIX atlas: <http://rdf.ebi.ac.uk/resource/atlas/>PREFIX atlasterms: <http://rdf.ebi.ac.uk/terms/atlas/>PREFIX up:<http://purl.uniprot.org/core/> PREFIX biopax3:<http://www.biopax.org/release/biopax-level3.owl#>SELECT distinct ?protein ?expressionValue ?pvalue WHERE {

?factor rdf:type efo:EFO_0000384 . ?value atlasterms:hasFactorValue ?factor . ?value atlasterms:isMeasurementOf ?probe . ?value atlasterms:pValue ?pvalue . ?value rdfs:label ?expressionValue . ?probe atlasterms:dbXref ?protein . FILTER ( ?pvalue < 1e-12 )FILTER ( strstarts(str(?protein),"http://purl.uniprot.org/uniprot/") )}

}ORDER BY ASC (?pvalue)

https://www.ebi.ac.uk/rdf/services/atlas/sparql

• Links pathways with genes, terms from Pathway, Cell line and Disease ontology, PubMed references

• Models individual Interactions• Can be downloaded as RDF• Has an experimental SPARQL endpoint

WikiPathways

• Define a query to find pathways linked to TNFalpha gene

Exercise

PREFIX wp: <http://vocabularies.wikipathways.org/wp#>PREFIX dc: <http://purl.org/dc/elements/1.1/>PREFIX dcterms: <http://purl.org/dc/terms/>PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>

SELECT DISTINCT ?PathwayName where {?geneProduct a wp:GeneProduct .?geneProduct dc:identifier ?GeneID .?geneProduct dcterms:isPartOf ?pathway . ?geneProduct rdfs:label ?geneName .?pathway dc:identifier ?pathwayid . ?pathway dc:title ?PathwayName . FILTER(str(?geneName) = "TNFalpha" )

}

http://sparql.wikipathways.org

• Try this, or another query– Using web interface– Using http get• Define a simple describe• Use a web tool to URLEncode the query• Submit query as a URL parameter

Exercise

DisGeNet

PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>PREFIX dcterms: <http://purl.org/dc/terms/>PREFIX foaf: <http://xmlns.com/foaf/0.1/>PREFIX skos: <http://www.w3.org/2004/02/skos/core#>PREFIX void: <http://rdfs.org/ns/void#>PREFIX sio: <http://semanticscience.org/resource/>PREFIX ncit: <http://ncicb.nci.nih.gov/xml/owl/EVS/Thesaurus.owl#>PREFIX up: <http://purl.uniprot.org/core/> SELECT DISTINCT ?gene WHERE {

?gda sio:SIO_000628 ?gene,?disease .?gene a ncit:C16612 . ?gene skos:exactMatch ?GeneID .?disease a ncit:C7057 .?disease dcterms:title ?DiseaseName .?gda sio:SIO_000216 ?scoreIRI .?scoreIRI sio:SIO_000300 ?score .FILTER (?score > "0.35"^^xsd:decimal) FILTER (contains(str(?DiseaseName),"Crohn"))

}

http://rdf.disgenet.org/lodestar

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>PREFIX owl: <http://www.w3.org/2002/07/owl#>PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>PREFIX dcterms: <http://purl.org/dc/terms/>PREFIX foaf: <http://xmlns.com/foaf/0.1/>PREFIX skos: <http://www.w3.org/2004/02/skos/core#>PREFIX void: <http://rdfs.org/ns/void#>PREFIX sio: <http://semanticscience.org/resource/>PREFIX ncit: <http://ncicb.nci.nih.gov/xml/owl/EVS/Thesaurus.owl#>PREFIX up: <http://purl.uniprot.org/core/>PREFIX wp: <http://vocabularies.wikipathways.org/wp#>PREFIX dc: <http://purl.org/dc/elements/1.1/>PREFIX dcterms: <http://purl.org/dc/terms/>

http://rdf.disgenet.org/lodestar

SELECT DISTINCT ?PathwayName WHERE {?gda sio:SIO_000628 ?gene, ?disease .?gene a ncit:C16612 .?disease a ncit:C7057 .?disease dcterms:title ?DiseaseName .?gda sio:SIO_000216 ?scoreIRI .?scoreIRI sio:SIO_000300 ?score .FILTER (?score > "0.35"^^xsd:decimal) FILTER (contains(str(?DiseaseName),"Crohn")) SERVICE <http://sparql.wikipathways.org/> {

?geneProduct a wp:GeneProduct .?geneProduct dc:identifier ?gene .?geneProduct dcterms:isPartOf ?pathway .?pathway dc:identifier ?pathwayid . ?pathway dc:title ?PathwayName .

} }

http://rdf.disgenet.org/lodestar/sparql

Bio ontologies and semantic technologies

Education