![HyBrowComputationally augmented method for hypothesis evaluationdeveloped by Racunas et al. [1]](https://image.slidesharecdn.com/tamaleseminarslideshare-101206104312-phpapp02/85/TAMALE-Seminar-Evaluating-scientific-hypotheses-using-Semantic-Web-technologies-14-320.jpg)
![incremental hypothesis improvement[1] Racunas S. A., Shah N. H., Albert I. and Fedoroff N. V. (2004). HyBrow: A prototype system for computer-aided hypothesis evaluation. Bioinformatics 20(S. 1): i1-i8.](https://image.slidesharecdn.com/tamaleseminarslideshare-101206104312-phpapp02/85/TAMALE-Seminar-Evaluating-scientific-hypotheses-using-Semantic-Web-technologies-20-320.jpg)
TAMALE Seminar: Evaluating scientific hypotheses using Semantic Web technologies
- 1. Evaluating scientific Hypotheses using semantic web technologiesAlison CallahanTAMALE SeminarNovember 8 2010
- 2. Why the semantic web?PLAIN TEXTWhat you see: “The weather today, November 8, will be cloudy with a high of 7°C”What your computer sees: akfalksjdfoaohwoiehroeXMLWhat you see:<weather> <date> <month>November</month> <day>8</day> <year>2010</year> </date> <temperature> <value>7</value> <unit>Celcius</unit> </temperature> <conditions>cloudy</conditions></weather>What your computer sees:<weather> <date> <month>November</month> <day>8</day> <year>2010</year> </date> <temperature> <value>5</value> <unit>Celcius</unit> </temperature> <condi>Flurries</condi></weather>
- 5. RDFResource Description Framework: A W3C standard for representing resources and the relationships between them, and for data exchange on the WWWAlisonreadinglikes subject predicate object http://people.com#Alisonhttp://feelings.com#likes http://activities.com#reading<rdf:RDF> <rdf:Descriptionrdf:about="http://people.com#Alison"><likesxmlns="http://feelings.com#"rdf:resource="http://activities.com#reading"/> </rdf:Description></rdf:RDF>
- 6. educational institutionfoaf:personrdf:typeCarleton Universityrdf:typeworks atreadingAlisonlikeshas brotherChrisrdf:typefoaf:person
- 7. Querying RDF using SPARQLSPARQL = SPARQL Protocol and RDF Query Language select ?s where { ?s rdf:typefoaf:Person . }http://people.com#Alisonhttp://people.com#Christopher
- 8. OWL: The web ontology languageOWL allows the representation of ontology concepts in a machine understandable mannerMotherWomanhasChildPerson
- 9. Biological SCIENCES and the semantic webhttp://bio2rdf.org http://bioportal.bioontology.org
- 10. HYQUEHypothesis-based query and evaluation toolhttp://semanticscience.org/projects/hyque
- 13. Finding evidence to support/refute a hypothesis is becoming increasingly difficultSource: http://upload.wikimedia.org/wikipedia/commons/2/26/EnwikipediaArt.PNG
- 14. HyBrowComputationally augmented method for hypothesis evaluationdeveloped by Racunas et al. [1]
- 16. uses consistency checking to evaluate hypotheses
- 17. constraints
- 18. rules
- 19. compares hypotheses using neighborhood functions
- 20. incremental hypothesis improvement[1] Racunas S. A., Shah N. H., Albert I. and Fedoroff N. V. (2004). HyBrow: A prototype system for computer-aided hypothesis evaluation. Bioinformatics 20(S. 1): i1-i8.
- 21. HyBrowsmall, manually generated knowledge
- 22. hard coded Perl rules
- 23. challenging to apply to a new domain
- 24. needs access to a greater KBHyBrow HyQueHypothesis query and evaluation system
- 26. Background knowledge encoded as OWL ontologies
- 27. Queries Bio2RDF’s dedicated SPARQL endpoints
- 28. Context-specific rules that consider experimental conditions
- 29. HyQueconsumes and producesRDFPaper: Callahan, A., M. Dumontier & N. Shah. 2010. HyQue: Evaluating hypotheses using Semantic Web technologies. Bio-ontologies SIG, ISMB’10, Boston MA.On the web: http://semanticscience.org/projects/hyque
- 30. HyQue is composed of …HyQue hypothesis ontologyDescribes generic input hypothesis and output hypothesis evaluation classesUses upper level classes e.g. ‘proposition’, ‘measurement value’, ‘event’ HyBrow SPARQL endpoint
- 31. SGD data in Bio2RDF
- 32. Template event-based SPARQL queries
- 33. GO, SO, ChEBI, ECO ontologiesA HyQue hypothesis is a collection of propositionsproposition: “a statement expressing something true or false”
- 34. HyQue hypotheses are composed of propositions connected using logical operators (AND, OR…)HyQuehypothesis ‘proposition’ that ‘has part’ some ‘hypothesis part’Hypothesis part ‘proposition’ that ‘has component’ some ‘event’
- 35. HyQue eventsEvents are composed of conditional assertions on a relation between ‘actor’ and ‘target’ induces(actor, target, context, location)For decidable logic (OWL), an n-ary object is usedEventagent_aactoragent_btargetperturbation_contextcontextphysical_locationlocation
- 36. HyQue Data … but first, a biology primer!
- 38. How are these processes regulated? Source: http://www.webcitation.org/5u4OelqJO
- 39. Where do OUR CELLS get energy?
- 40. HyQue data: The gal gene network IN YEASTGenes that encode proteins that transport and metabolize galactose
- 41. permease – gal2p – transports galactose into cells
- 46. Regulation – whether the pathway is on or off
- 47. gal3p
- 48. gal4p
- 49. gal80penzymes that use galactose
- 50. Source: Ostergaardet al. (2000). Nature Biotechnology 18: 1283 - 1286
- 51. HyQue dataExperimentally determined interactions between the GAL proteinsProperties of the genes that encode these proteins (SGD)Literature-based evidence (citations)Knowledge about cellular locations and events (GO)Types of evidence supporting these interactions (ECO)
- 52. How does HyQue work?
- 53. Individual events parsed from input hypothesis RDFhypothesis:h a hyque:Hypothesis ; a hyque:AND ;hyque:hasPart :p1 ;hyque:hasPart :p2 ;hyque:hasPart :p3 .:p1 a hyque:AND ;hyque:hasComponent :e1 ;hyque:hasComponent :e2 .:e1 a <http://bio2rdf.org/go:0006810> ;hybrow:is_negated “0”^^xsd:boolean ;hybrow:agent_a <http://bio2rdf.org/sgd:Gal2p> ;hybrow:agent_b <http://bio2rdf.org/chebi:28260> .:e2 a <http://bio2rdf.org/go:0005488> ;hybrow:is_negated “0”^^xsd:boolean ;hybrow:agent_a <http://bio2rdf.org/sgd:Gal3p> ;hybrow:agent_b <http://bio2rdf.org/sgd:Gal80p> .has parthypothesis part 1has componentgal2p transports galactosegal3p binds to gal80p
- 54. Template SPARQL queries completed based on event properties:e1 a go:0006810 ;hybrow:is_negated "0" ;hybrow:agent_asgd:Gal2p ;hybrow:agent_bchebi:28260 .construct { … } where {… ?event hybrow:is_negated ?negated .?event hybrow:physical_operator ?physical_operator .?event hybrow:agent_a <http://bio2rdf.org/sgd:Gal2p> .…?event hybrow:agent_b <http://bio2rdf.org/chebi:28260> .?actor semsci:isLocatedIn ?actor_gp_id_location .?actor_gp_id_locationrdf:type ?actor_location_type .?actor semsci:hasFunction ?actor_gp_id_function .?actor_gp_id_functionrdf:type ?actor_function .…}
- 55. SPARQL query results retrievedhybrow_data:2c1789a3019fd2fe9843d507824fc591 rdf:type <http://bio2rdf.org/go:0044092> .hybrow:is_negated "0" .hybrow:agent_a sgd:Gal3p ;hybrow:agent_b sgd:Gal80p ;hybrow:actor_type <http://bio2rdf.org/chebi:36080> ;hybrow:target_type <http://bio2rdf.org/chebi:36080> ;hybrow:physical_context <http://bio2rdf.org/go:0005634> ;hybrow:physical_operator <http://bio2rdf.org/go:0005488> .hybrow_data:b09f7cc043201b47610c874499448a23 rdf:type <http://bio2rdf.org/go:0005488> ;hybrow:is_negated "0" ;hybrow:agent_a sgd:Gal3p ;hybrow:agent_b sgd:Gal80p ;hybrow:actor_type <http://bio2rdf.org/chebi:36080> ;hybrow:target_type <http://bio2rdf.org/chebi:36080> ;hybrow:physical_context <http://bio2rdf.org/go:0005634> ;hybrow:physical_operator <http://bio2rdf.org/go:0005488> .
- 56. Query results evaluated based on rule sets‘binding’ rule:Is event negated?If yes, subtract 2Is physical operator ‘binding’?If yes, add 1; if no, subtract 1Is actor of type ‘protein’ or ‘small molecule’?If yes, add 1; if of type ‘gene’, subtract 1Is target of type ‘protein’ or ‘small molecule’? If yes, add 1; if of type ‘gene’, subtract 1Does actor have known ‘binding’ function? If yes, add 1 GO:0005488CHEBI:36080SO:0000236
- 57. Result scores based on operators between eventsFinal score = e1 score + e2 score + e3 score + e4 scoreFinal score = maximum of e5, e6 or e7 scores:p1 a hyque:AND ;hyque:hasComponent :e1 ;hyque:hasComponent :e2 ; hyque:hasComponent :e3 ;hyque:hasComponent :e4 .:p2 a hyque:OR ;hyque:hasComponent :e5 ;hyque:hasComponent :e6 ;hyque:hasComponent :e7 .
- 58. Hypothesis evaluation RDF generated
- 61. HyQue as a SADI serviceSADI – Semantic Automated Discovery and IntegrationHyQueSADI ontology to describe input and output of service
- 62. Users can post a hypothesis in RDF and receive the hypothesis evaluation in RDFACKNOWLEDGEMENTSMy supervisor: Dr. Michel DumontierProject collaborator: Dr. Nigam Shah http://stanford.edu/~nigamThe Dumontier Lab: http://dumontierlab.comFunding: NSERC to MD
- 63. Questions?