An Analysis of the Orphanet Ontology of Rare Diseases, Paris, October 2012

I presented some of my work to researchers at INSERM/ICS in Paris, who were working on the Orphanet Ontology of Rare Diseases. It's a very broad overview of my work on justificatory structure, justification isomorphism, and Chiara Del Vescovo's work on the atomic decomposition of OWL ontologies.

Transcript

1. Samantha Bail The School of Computer Science The University of Manchester, UK An Analysis of the Orphanet Ontology of Rare Diseases Paris, 18/10/2012
2. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Our current OWL research 2 •Justificatory structure (explanation) - Samantha Bail •Ontology diff & hotspot detection - Rafael Goncalves •Atomic decomposition (modularity) - Chiara Del Vescovo •Patterns & regularities - Eleni Mikroyannidi •Our aims: ‣ Gain a better understanding of OWL ontologies ‣ Provide OWL ontology developers with ‣ analytical tools ‣ development support
3. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Justificatory structure: Background 3 inference rules ontology (set of axioms) entailment (axiom) A justification is a minimal subset of the ontology which is sufficient for an entailment to hold. Definition
4. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Justificatory structure: Background 4 inference rules justi cation (set of axioms) entailment (axiom) A justification is a minimal subset of the ontology which is sufficient for an entailment to hold. Definition
5. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Justificatory structure: Background 5 entailment (axiom) A justification is a minimal subset of the ontology which is sufficient for an entailment to hold. Definition
9. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases ‘Strict’ isomorphism •Isomorphism [1] between justifications is well defined •It describes an equivalence relation 9 [1] Matthew Horridge, Samantha Bail, Bijan Parsia, and Ulrike Sattler. The cognitive complexity of OWL justifications. In Proceedings of ISWC-11, 2011. J1 = {A ￿ B ￿ ∃r.C, B ￿ ∃r.C ￿ D} |= A ￿ D J2 = {E ￿ B ￿ ∃s.F, B ￿ ∃s.F ￿ D} |= E ￿ D φ = {A ￿→ E, C ￿→ F, r ￿→ s} Example
10. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Subexpression-isomorphism •Covers justifications that have different subexpressions •Semantics of subexpressions do not matter 10 Example J1 = {A ￿ B ￿ C, B ￿ C ￿ D} |= A ￿ D J2 = {A ￿ ∃r.C, ∃r.C ￿ D} |= A ￿ D
11. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Subexpression-isomorphism •Covers justifications that have different subexpressions •Semantics of subexpressions do not matter 11 Example J1 = {A ￿ B ￿ C, B ￿ C ￿ D} |= A ￿ D J2 = {A ￿ ∃r.C, ∃r.C ￿ D} |= A ￿ D X1
12. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Subexpression-isomorphism •Covers justifications that have different subexpressions •Semantics of subexpressions do not matter 12 Example J1 = {A ￿ B ￿ C, B ￿ C ￿ D} |= A ￿ D J2 = {A ￿ ∃r.C, ∃r.C ￿ D} |= A ￿ D X1 X1
13. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Subexpression-isomorphism •Covers justifications that have different subexpressions •Semantics of subexpressions do not matter 13 Example J1 = {A ￿ B ￿ C, B ￿ C ￿ D} |= A ￿ D J2 = {A ￿ ∃r.C, ∃r.C ￿ D} |= A ￿ D X1 X2 X1
14. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Subexpression-isomorphism •Covers justifications that have different subexpressions •Semantics of subexpressions do not matter 14 Example J1 = {A ￿ B ￿ C, B ￿ C ￿ D} |= A ￿ D J2 = {A ￿ ∃r.C, ∃r.C ￿ D} |= A ￿ D X1 X2 X1 X2
15. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Lemma-isomorphism •Covers justifications of expressions and different size •We restrict lemmas [2] to atomic subsumption chains 15 [2] Matthew Horridge, Bijan Parsia, and Ulrike Sattler. Lemmas for justifications in OWL. In Proceedings of DL-09, 2009. Example J1 = {A ￿ B, B ￿ C} |= A ￿ C J2 = {A ￿ B, B ￿ C, C ￿ D} |= A ￿ D
16. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Lemma-isomorphism •Covers justifications of different expressions and different size •We restrict lemmas [2] to atomic subsumption chains 16 [2] Matthew Horridge, Bijan Parsia, and Ulrike Sattler. Lemmas for justifications in OWL. In Proceedings of DL-09, 2009. Example J1 = {A ￿ B, B ￿ C} |= A ￿ C J2 = {A ￿ B, B ￿ C, C ￿ D} |= A ￿ D ...
17. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Analysis: BioPortal & OntoOrpha •Tested: •Jan 2012 snapshot of BioPortal •83 OWL and OBO ontologies •June 2011 version of OntoOrpha (lite_2011-06-17) •Terminology: •“non-trivial entailment”: entailment which has some justification other than the axiom itself •“buckets”: set of isomorphic justifications 17
18. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Analysis: Basic justification metrics •OntoOrpha metrics: •69387 logical axioms (avg: 4895) •Expressivity: ALEH •199* non-trivial entailments (avg: 2738) •1139 justifications for these entailments (avg: 78) •3.5 justifications per entailment (avg: 3.5) •2 axioms per justification (avg: 2.6) 18 * Justification generation timed out on some entailments
19. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Analysis: Basic justification metrics 19
20. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases OntoOrpha: Isomorphism overview 20
21. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases OntoOrpha: Frequent justification patterns 21 J1 = {sign id x ￿ ∃ frequentSignOf.pat id y, domain(frequentSignOf, clinical sign)} |= sign id x ￿ clinical sign x 901 x 228 J3 = {pat id a ￿ pat id b, pat id b ￿ pat id c} |= pat id a ￿ pat id c x 10 •Lemma-isomorphism reduces 1139 justifications to 3 patterns: J2 = {sign id x ￿ ∃ frequentSignOf.pat id y, frequentSignOf ￿ signOf domain(signOf, clinical sign)} |= sign id x ￿ clinical sign
22. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases OntoOrpha: Frequent justification patterns 22 •Domain axioms for frequentSignOf, veryFrequentSignOf, occasionalSignOf ‘trigger’ multiple justifications •Pattern J2 occurs only in OntoOrpha (out of 83 ontologies) J1 = {sign id x ￿ ∃ frequentSignOf.pat id y, domain(frequentSignOf, clinical sign)} |= sign id x ￿ clinical sign J2 = {sign id x ￿ ∃ frequentSignOf.pat id y, frequentSignOf ￿ signOf domain(signOf, clinical sign)} |= sign id x ￿ clinical sign
23. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases OntoOrpha: Frequent justification patterns 23 a63 j1 a50 j6 a51 j10 a0 j4 j5 j2 j3 j8 j9 j7 a16 a15 a12 a13 a10 a1 a5 a6 a56 a8 e0 •Part of a j-graph for one such entailment (pattern J2) •a0: domain(signOf, clinical sign) •a5, a6, a12: subproperty axioms
24. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases BioPortal: Pattern distribution 24 OntoOrpha
25. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition •Fine-grained type of modularity •Atom = set of terms which contains all the information related to a specific term •DeMoSt: Tool for exploring atomic decomposition and modular structure of an ontology 25
26. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Example 26 Pizza DisjointWith PizzaTopping Pizza SubClassOf Food Pizza SubClassOf hasBase some PizzaBase Pizza DisjointWith PizzaBase PizzaBase SubClassOf Food ThinAndCrispyBase SubClassOf PizzaBase CheeseTopping SubClassOf PizzaTopping Spiciness EquivalentTo Hot or Medium or Mild PeperonataTopping SubClassOf PepperTopping PizzaTopping SubClassOf Food Functional: hasSpiciness HamTopping SubClassOf MeatTopping Margherita SubClassOf hasTopping only (MozzarellaTopping or TomatoTopping) MushroomTopping SubClassOf VegetableTopping GarlicTopping SubClassOf hasSpiciness some Medium IceCream DisjointWith Pizza DeepPanBase DisjointWith ThinAndCrispyBase Capricciosa DisjointWith FourSeason RealItalianPizza SubClassOf hasBase only ThinAndCrispyBase PrinceCarlo SubClassOf hasTopping some ParmesanTopping AsparagusTopping DisjointWith TomatoTopping ChickenTopping DisjointWith HotSpicedBeefTopping
27. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Example 27 Pizza DisjointWith PizzaTopping Pizza SubClassOf Food Pizza SubClassOf hasBase some PizzaBase Pizza DisjointWith PizzaBase PizzaBase SubClassOf Food ThinAndCrispyBase SubClassOf PizzaBase CheeseTopping SubClassOf PizzaTopping Spiciness EquivalentTo Hot or Medium or Mild PeperonataTopping SubClassOf PepperTopping PizzaTopping SubClassOf Food Functional: hasSpiciness HamTopping SubClassOf MeatTopping Margherita SubClassOf hasTopping only (MozzarellaTopping or TomatoTopping) MushroomTopping SubClassOf VegetableTopping GarlicTopping SubClassOf hasSpiciness some Medium IceCream DisjointWith Pizza DeepPanBase DisjointWith ThinAndCrispyBase Capricciosa DisjointWith FourSeason RealItalianPizza SubClassOf hasBase only ThinAndCrispyBase PrinceCarlo SubClassOf hasTopping some ParmesanTopping AsparagusTopping DisjointWith TomatoTopping ChickenTopping DisjointWith HotSpicedBeefTopping Tell me everything about ‘Pizza’!
28. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Example 28 PizzaBase SubClassOf Food ThinAndCrispyBase SubClassOf PizzaBase CheeseTopping SubClassOf PizzaTopping Spiciness EquivalentTo Hot or Medium or Mild PeperonataTopping SubClassOf PepperTopping PizzaTopping SubClassOf Food Functional: hasSpiciness HamTopping SubClassOf MeatTopping Margherita SubClassOf hasTopping only (MozzarellaTopping or TomatoTopping) MushroomTopping SubClassOf VegetableTopping GarlicTopping SubClassOf hasSpiciness some Medium IceCream DisjointWith Pizza DeepPanBase DisjointWith ThinAndCrispyBase Capricciosa DisjointWith FourSeason RealItalianPizza SubClassOf hasBase only ThinAndCrispyBase PrinceCarlo SubClassOf hasTopping some ParmesanTopping AsparagusTopping DisjointWith TomatoTopping ChickenTopping DisjointWith HotSpicedBeefTopping Pizza DisjointWith PizzaTopping Pizza SubClassOf Food Pizza SubClassOf hasBase some PizzaBase Pizza DisjointWith PizzaBase
29. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Influence diagram 29 Pizza
30. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Influence diagram 30 Pizza PizzaBase PizzaTopping Food Influences
31. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Influence diagram 31 Pizza PizzaBase PizzaTopping Food NamedPizza Siciliana Margherita ... Influences
32. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Influence diagram 32 PizzaBase PizzaTopping Food NamedPizza Siciliana Margherita ... Influences Pizza IceCream
33. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Atomic decomposition: Influence diagram 33 PizzaBase PizzaTopping Food NamedPizza Siciliana Margherita ... Influences Pizza IceCream What does this tell us about the modelling in the ontology? Should ‘IceCream’ really be influenced by ‘Pizza’? Maybe this isn’t quite right yet...
34. Atomic decomposition of bottom module for gen_id_15998, gen_id_15999, gen_id_16000, gen_id_16001 selected atom affects meaning of selected atom not selected/affected
35. Atomic decomposition of bottom module for gen_id_15998, gen_id_15999, gen_id_16000, gen_id_16001 selected atom affects meaning of selected atom not selected/affected
36. influences
37. Electron-transferring-flavoprotein dehydrogenase, Glutaric acidemia type 2, geneOf, gene Fatty acid oxidation and ketogenesis disorder with hypertrophic cardiomyopathy componentPartOf Fatty acid oxidation or ketogenesis disorder Electron-transferring-flavoprotein dehydrogenase disease Lysine or tryptophan metabolism disorder Branched-chain amino acid metabolism disorder Muscular lipidosis influences
38. Samantha Bail An Analysis of the Orphanet Ontology of Rare Diseases Summary & conclusions 38 •Justification patterns differ strongly between ontologies •Regularities do not correlate with size or expressivity •Justifications in OntoOrpha are highly interconnected... •multiple justifications per entailment •shared domain axioms •... and very regular: only 3 patterns for 1139 justifications •... and quite similar to other ontologies •DeMoSt: There are some fairly large atoms... •... might be worth looking at them in more detail?
39. Justifications: bails@cs.man.ac.uk Atomic decomposition: delvescc@cs.man.ac.uk http://owl.cs.manchester.ac.uk Twitter: @semwebmcr Any questions? Thank you!