The document discusses representing and reasoning over chemical structures using OWL and description graphs, highlighting the challenges of maintaining structural complexity within ontology frameworks. It illustrates how a chemical ontology can be constructed with defined classes and properties, the application of rules for logical deductions, and the necessity of a strong separation requirement for maintaining decidability. The results indicate that while OWL can represent structures effectively, scalability and reasoning performance remain challenging as the knowledge base grows.

1. Representing chemicalsusing OWL, Description Graphsand RulesJanna Hastings, EBI, UKMichel Dumontier, Carleton University, CanadaDuncan Hull, EBI, UKMatthew Horridge, Manchester, UKChristoph Steinbeck, EBI, UKUlrike Sattler, Manchester, UKRobert Stevens, Manchester, UKTertia Hӧrne, University of South AfricaKatarina Britz, Meraka Institute, South AfricaOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

2. OWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

3. ProblemWe wish to represent and reason over structured objects i.e. their representation contains also their partsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

4. Chemical structuressingle bondCarbon atomHydrogen atomdouble bondNitrogen atomOxygen atomcaffeineMoleculesconsist of atomsconnected by bondsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

5. Chemical ontologyChemical ontology consists of chemical classes which can be defined by parts of structuresand/or properties of structurescarboxylic acidif molecule has partsome carboxy groupcyclic moleculeif molecule has property cyclic, i.e. a self-connectedcyclic path exists through the molecule’s atomsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

6. OWL representationWithout structure, all parts must be explicitly asserted(combinatorial explosion for larger molecules)But the structure of complex molecules breaks the OWL Tree Model requirementdoes not have a model in the shape of a treeOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

7. Description GraphsA recent, decidable extension to OWL 2, allowing expression of complex structures as graphs within the ontologyA description graph consists of a set of labelled vertices and a set of directed edgesEach description graph has a main class which links the graph to the main OWL ontologyOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

8. Strong separationIn order to preserve decidability of knowledge bases enriched with description graphs, atomic properties used as graph edges have to be different to those used in axioms in the main OWL ontologyThis is known as the strong separation requirementOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

9. RulesEnhance OWL with the capacity to express if – then constructionsConsist of ‘antecedent’ (if conditions) and ‘consequent’ (then result)Antecedent and consequent are composed of conjunctions of atomic statementsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

10. GoalCan we represent chemical structures using OWL and Description Graphs?Can we reason over the information encoded in chemical structures using OWL, Description Graphs and Rules?OWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

11. OWL ontologyOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

12. Chemical description graphsGenerated based on structures converted from a chemical database (ChEBI)OWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

13. RulesGenerated for properties, e.g. being cyclicmolecule(?x), atom(?a1), atom (?a2), atom(?a3), atom(?a4), bond(?b1), bond (?b2), bond(?b3), bond (?b4), has_atom(?x, ?a1), has_atom(?x, ?a2), has_atom(?x, ?a3), has_atom(?x, ?a4),has_bond(?a1, ?b1), has_bond(?a1, ?b4), has_bond(?a2, ?b1), has_bond(?a2, ?b2),has_bond(?a3, ?b2), has_bond(?a3, ?b3), has_bond(?a4, ?b3), has_bond(?a4, ?b4)-> cyclic_entity(?x)cyclobutanetetrahedraneOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

14. RulesGenerated for parthood, e.g. carboxylic acidmolecule(?y), atom(?a0), oxygen_atom(?a1), carbon_atom(?a2), oxygen_atom (?a3), has_atom(?y, ?a0), has_atom (?y, ?a1), has_atom (?y, ?a2), has_atom (?y, ?a3), double_bond(?b0), single_bond (?b1), single_bond (?b2), has_bond(?a0, ?b2), has_bond(?a1, ?b1), has_bond(?a2, ?b0), has_bond(?a2, ?b1), has_bond(?a2, ?b2), has_bond(?a3, ?b0) -> carboxylic_acid(?y)carboxylic acidbenzoicacidbenzoic acid has this part so: is a carboxylic acidOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

15. Testing the reasoningCan we use a reasoner to deduce the classification hierarchy based on the graphs and rules? No asserted hierarchy between test classes and molecules with generated graphsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

16. ResultsInferred hierarchy shows classified moleculesOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

17. Testing the performanceHow many molecules (description graphs) can we include in our knowledge base?How does the reasoning task (classification) scale with respect to the number of graphs, both with and without rules in the knowledge base?OWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

18. ResultsOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

19. ExperiencesDifficult to debugTools support needs to be improvedDifficult to construct rules for properties which depend on all atoms or all bonds in a given molecule e.g. saturated -> all bonds in molecule are singleOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

20. ConclusionUsing OWL, Description Graphs and Rules we can represent chemical structures at the class level in our knowledge base and reason over the structural informationScalability of the reasoning with the rules is a concernOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules

21. AcknowledgementsSpecial thanks to KirillDegtyarenko, Stefan Schulz, Colin Batchelor, BirteGlimm and the ChEBI teamFunding: Meraka Institute, South Africa; BBSRC (BB/G022747/1); NSERC Discovery GrantOWLED2010:San Francisco:Representing chemicals using OWL, Description Graphs and Rules