Workshop on Ontologies and their Applications September 28

Workshop on Ontologies and their Applications, September 28, 2004, São Luís do Maranhão (Brazil) Ontological Foundations for Biomedical Sciences Stefan Schulz, Kornél Markó, Udo Hahn Department of Medical Informatics University Hospital Freiburg (Germany) Text Knowledge Engineering Lab University of Jena (Germany)

The World of Life Sciences… Evolution of Life Millions of Species Morphology Function …requires sophisticated organization Dysfunction Organisms Organ Systems Organs Tissues Cells Genes Molecules

Ontologies Domain Philosophy Computer Science Cognitive Science Logics

Bio-Ontologies Domain: Biology Computer Science Philosophy: Realism Logics: Cognitive Science FOL

Top-Level Division Biological Entitiy Continuants: Occurrents: (Changes of) states of affairs of the physical world: Examples: process, state, event, disease, procedure… Entities of the physical world („Biomedical Structure“): Examples: body, organ, tissue, molecule, . . disjoint depend on

Representation of Continuants in Bio-ontologies. What exists ? n Human Anatomy n Foundational Model of Anatomy (FMA) n Portions of SNOMED, Open. Galen, Me. SH n Other Organisms n Open Biological Ontologies (OBO) n. Mouse (developmental stages), Zebrafish, Drosophila, … n UMLS Semantic Network n Species-Independent n Gene Ontology: Cellular Component branch Size: 14 (UMLS SN) – 103 (Adult Mouse) – 105 (FMA)

Deficiencies of existing Bio. Ontologies n. Redundancy n. Synonymy n. Ambiguity n. Underspecification

Redundancy

Synonymy n “Motor Neuron instance-of Neuron” (Fly. Base) n “Motor Neuron narrower Neuron” (Me. SH) n “Motor Neuron subclass-of Neuron” (FMA, Open. GALEN) Neuron Is-A ? Motor Neuron

Ambiguity, Underspecification n “Cell has-part Axon” (Gene Ontology) n Do cells without axons exist ? n Do axons without cells exist ? n “Neuron has-part Axon” (FMA) n Does every neuron has an axon?

Ambiguity, Underspecification n “Cell has-part Axon” (Gene Ontology) n Do cells without axons exist ? n Do axons without cells exist ? n “Neuron has-part Axon” (FMA) n Does every neuron has an axon? “Keep in mind that part_of means can be a part of, not is always a part of “ GO Editorial Style Guide, Oct 2003 “The part_of relationship (…) is usually necessarily is_part” GO Editorial Style Guide, Jan 2004 “A part_of B if and only if: for any instance x of A there is some instance y of B which is such that x stands to y in the instance-level part relation, and vice versa”. Rosse & Smith MEDINFO 2004

Semantic framework for biological structure… n. Foundational Relations n. General Attributes n. Theories

Bio-ontologies Occurrents Continuants Life, Appendectomy, Mitosis Hand, Blood, Cell, Tree My Life, Appedectomy of Patient #123, this Mitosis My Hand, Blood Sample #12345, this Cell, the Maple Tree in front of the house #xyz (Changes of) states of affairs Entities of the physical world: Universals (Concepts, Classes of Individuals) Particulars (Concrete Objects in the world) Four disjoint partitions

Some Foundational Relations between Biological Continuants Rel(x, y) y x Universals Particulars Is-A Instance-of part-of, has-location has-branch, bounds, connects has-developmental-form

Some Foundational Relations between Biological Continuants Rel(x, y) y x Universals Particulars Is-A, Part-Of, Has-Location Bounds, Has-Branch, Connects Has-Developmental-Form Instance-of part-of, has-location has-branch, bounds, connects has-developmental-form

From Instance-to-Instance relations to Class-to-Class Relations A, B are classes, membership rel: relation between instances Rel (A, B) =def inst-of = class Rel: relation between classes x: inst-of (x, A) inst-of (y, B) rel (x, y) OR x: inst-of(x, A) y: inst -of (y, B) rel (x, y) OR y: inst-of(y, B) x: inst-of (x, A) rel (x, y) cf. Schulz & Hahn (KR 2004, ECAI 2004) Rosse & Smith (MEDINFO 2004)

From Instance-to-Instance relations to Class-to-Class Relations A, B are classes, membership rel: relation between instances Rel (A, B) =def inst-of = class Rel: relation between classes x: inst-of (x, A) inst-of (y, B) rel (x, y) OR x: inst-of(x, A) y: inst -of (y, B) rel (x, y) AND y: inst-of(y, B) x: inst-of (x, A) rel (x, y) cf. Schulz & Hahn (KR 2004, ECAI 2004) Rosse & Smith (MEDINFO 2004)

Semantic framework for biological structure… n. Foundational Relations n. General Attributes n. Theories

General Attributes (mutually disjoint classes) n Dimensionality: Point, 1 -D, 2 -D, 3 -D

General Attributes (mutually disjoint classes) n Dimensionality: Point, 1 -D, 2 -D, 3 -D n Solids vs. hollow spaces, vs. Boundaries

General Attributes (mutually disjoint classes) n Dimensionality: Point, 1 -D, 2 -D, 3 -D n Solids vs. hollow spaces, vs. Boundaries n Collections vs. Masses vs. Count Objects cf. Schulz & Hahn, FOIS 01

Semantic framework for biological structure… n. Foundational Relations n. General Attributes n. Theories

Theories n A set of formal axioms which describe a restricted (local) domain. n Four orthogonal theories for Biological Structure n. Granularity n. Species n. Development n. Canonicity

Granularity n Level of detail (molecular, cellular, tissue, organ) n Change in Granularity level may be nonmonotonous n Change of sortal restrictions: n 3 -D 2 -D boundary n. Count concept Mass concept n Change of relational attributions: ndisconnected

Theories n A set of formal axioms which describe a restricted (local) domain. n Four orthogonal theories for Biological Structure n. Granularity n. Species n. Development n. Canonicity

Linnean Taxonomy of Species http: //tolweb. org

Species Introduction of axioms at the highest common level Has-Part Skull Has-Part Vertebra Has-Part Jaw

Theories n A set of formal axioms which describe a restricted (local) domain. n Four orthogonal theories for Biological Structure n. Granularity n. Species n. Development n. Canonicity

Development n Represents timedependent “snapshots” from the life cycle of an organism, e. g. , zygote, embryo, fetus, child, adult n Development stages are speciesdependent e. g. metamorphosis

Theories n A set of formal axioms which describe a restricted (local) domain. n Four orthogonal theories for Biological Structure n. Granularity n. Species n. Development n. Canonicity

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure acquired congenital

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure n Lethal Structure

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure n Lethal Structure n Derivates of biological structure

Canonicity n Five canonicity levels: each level introduces axioms valid for higher levels

Examples Granularity Species Development Canonicity low high general specific embryo adult low high

Coverage: Foundational Model of Anatomy Granularity Species Development Canonicity low high general specific embryo adult low high

Coverage: Gene Ontology Granularity Species Development Canonicity low high general specific embryo adult low high

Coverage: Mouse Anatomy Granularity Species Development Canonicity low high general specific embryo adult low high

Examples Connects (Right. Ventricle, Left Ventricle) Granularity Species Development Canonicity = normal = mammal = adult = 4 -5 Granularity Species Development Canonicity = any = vertebrate = early embryo = any false true Is-A (Membrane, 3 -D object) Granularity Species Development Canonicity = normal = any true Granularity Species Development Canonicity = lowest = any false

Conclusion n Integration of bio-ontologies requires n Uncontroversial semantics of relations and attributes n Clear commitment to theories, such as granularity, species, development and canonicity n Redundancy can be avoided n Encoding axioms at the highest common level in the species taxonomy (e. g. vertebrates, arthropods, primates) and benefit from inheritance in subsumption hierarchies

…requires sophisticated organization n Formalization and Standardization of Clinical Terminologies n Basis for the Annotation of Genes and Gene Products n Semantic reference for scientific communication n Machine-supported reasoning and decisionsupport Bio-ontologies !

Upper level classification of entities Individuals (concrete objects) • my left hand Continuants • a blood sample (physical objects, …) • a concrete cell Occurrents (events, processes, actions…) • Peter’s diabetes • appendectomy of Patient #12345 Universals (Concepts, Classes of Individuals) • Hand, • Blood • Cell • Diabetes mellitus • Appendectomy

Mereotopological Quiz • Cranial Cavity has-location Head Is Cranial Cavity part-of Head ? • Brain has-location Cranial Cavity Is Brain part of Cranial Cavity ? • Glioblastoma has-location Brain Glioblastoma part-of Brain? • Brain metastasis has-location Brain metastasis part-of Brain? • Embryo has-location Uterus Embryo part-of Uterus ? Images from: Sobotta CD-ROM

part-of HL has-location is-a HS (Solid) is HE Head -a -a is Head is-a HP p SL is-a has-location SS is-a is -a (Solid) SE Skull -a is Skull SP p CL is-a has-location -a is is-a p -a (Hole) CE Cranial Cavity is Cranial Cavity CS CP BL has-location is-a is BE Brain p -a (Solid) is Brain -a BS BP transitive closure by taxonomic subsumption has-location

Subtheories of an Ontology of Biological Structure 1. Taxonomy „is-a“ 2. 3. Topology „part-of“ „connection“ Mereology Hand part-of Thumbnail

Subtheories of an Ontology of Biological Structure 1. Taxonomy „is-a“ 2. Mereology „part-of“ 3. Topology „connection“ • Canonical relationships X Y X Y XY (Schulz et al. AMIA 2000) • Topological Primitives: X Y Y X XY YX

Structure of Talk n Introduction n Foundational Relations n Foundational Attributes n Theories n Granularity n Species n Development n “Canonicity”

The World of Life Sciences…

Generalized Representation of Living Systems: Top Level Biological Entities Biological Occurrents process, state, event, … dependence Biological Continuants organism, organ, tissue, cell, molecule, . .

Ontological Account for Biological Continuants n. Foundational Relations n. Foundational Attributes n. Theories n. Granularity n. Species n. Development n“Canonicity”

Granularity n Taxonomic: degree of specialization n Mereologic: degree of dissection {molecular level, cellular level, tissue level, organ level, population level}

Change in Granularity level may be non-monotonous n Change of sortal restrictions: n 3 -D 2 -D boundary n Count concept Mass concept n Change of relational attributions: n disconnected n

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure n Lethal Structure

Canonicity n Degrees of “Wellformedness” of Biological Structure: n Canonic structure n Structural Variations n Pathological Structure n Lethal Structure n Derivates of biological structure