What ontologies exist, who builds them and and what are they used for?

An overview of the ontology landscapeAn overview of the ontology landscape

Robert Stevens, James Malone

robert.stevens@manchester.ac.uk, malone@ebi.ac.uk

Outline

• What do we need to describe?

• What exists to describe it?

• Are they any good….?

• Ontology organisations

Dimensions of description

• The entities themselves – genes proteins, processes, cells, properties

• The investigations that produced the entities

• The informational origins and history of those entities and their descriptions (data and provenance)

What entities exist to be described

• The actual “concrete” biological entities themselves: Proteins, genes, small molecules, cells, gross anatomy, etc etc

• The devices used to produce and measure them

• Properties of those entities: Size, Shape, colour, function, role, etc etc.

• The biological processes in which those biological entities take part.

• The measuring and analytical processes used on those biological entities.

• Sites on those biological entities: Shoulder region, a bit of the environment, the dorsal region of a mouse, etc etc.

• Information artefacts about all of the above: sequences, database records, who, what, when, where and how… lab protocols, etc etc.

Dividing things up from the top

Dividing things up from the top - process

Gene ontology (GO)biological process,Gene ontology (GO) molecular process

Dividing things up from the top - information

Information Artifact Ontology (IAO)Software Ontology (SWO)Unit Ontology (UO)

Dividing things up from the top - material

ChEBIProtein Ontology (PrO)Sequence Ontology (SO)Cell Type Ontology (CLO)Uberon Foundational Model of Anatomy (FMA)NCBI Taxonomy

Dividing things up from the top - property

GO Molecular FunctionPhenotypic Quality (PaTO)Human Disease Ontology (HDO)

Dividing things up from the top - site

Gazetteer Ontology (GAZ)

We’ve covered most of what there is…

• We’ve chosen bits from a simple upper level ontology

• These are domain neutral descriptions of the entities in any domain of interest

• Top-level or upper ontologies give a common view on what discriminations to make…

• … and what relationships to use between them

• BFO, Simple top Bio

Ontologies in these dimensions

• Here we want a “space” covering these dimensions with ontologies splattered about

• Dimension 1: genotype to phenotype

• Dimension 2: investigations

• Dimension 3: information – IAO, prov, etc.

Reference vs Application Ontologies

• Ontologies developed for different uses

• Reference ontologies built with aim of becoming authority on given domain

• Application ontologies built towards specific application use cases, such as for tooling or database needs

• Application ontologies often consume reference ontologies

Things we describe in Biology - Genes

• Gene Ontology - Gene biological processes, cellular components and molecular functions

• Seen as benchmark of success in bio-ontology

• Many ‘best practices’ fallen out of the GO’s development such as evidence codes, obsolescence policy and community development

Things we describe in Biology - Phenotypes• PATO – ‘phenotypic qualities’, i.e. physical properties of

organisms

• Extremely wide range of classes, examples include colour, size, shape, odour, behaviour

• Phenotypes are important in understanding how genes interact with the environment (in producing phenotypes)

Matzke MA, Image: Matzke AJM (2004) Planting the Seeds of a New Paradigm. PLoS Biol 2(5): e133

Master headline

Things we describe in Biology - Disease

• Majority of biomedical studies consider disease in some way

• Multiple terminologies for disease on biology

• SNOMED CT – Medical (clinical) terminology

• ICD-10 – Classification of disease and health problems

• NCI Thesaurus (not an ontology) - large, lots of textual definitions but less axiomatisation, disease subpart

• UMLS – set of controlled vocabularies describing medical concepts very large at >1 million biomedical concepts

• Human Disease Ontology – based on subset of UMLS, enriched with relationships and new concepts

Master headline

Things we describe in Biology - Anatomy

• Anatomy is important for many reasons including:

• Understanding how genes relate to anatomical regions

• Understanding how disease affects anatomical systems

• Comparative anatomy, i.e. comparing how structures in different species are related

• Model organism anatomies, e.g.

• Mouse adult gross anatomy

• Human anatomy – FMA

• Drosophila Anatomy

• Arabidopsis thaliana

• Zebrafish

• C. elegans

Master headline

Genes at work in different species anatomy • DII gene orthologs implicated in development in multiple

species of different anatomical parts

Mungall, C. et al (2012) Uberon, an integrative multi-species anatomy ontology. Genome Biology 2012, 13:R5

Things we describe in Biology – Chemical Entities

• ChEBI - molecular entities focused on ‘small’ chemical compounds

• Janna will talk about this tomorrow

Things we describe in Biology – Cells

• Cell Ontology is an ontology of cell types

• CL merges information contained in species-specific anatomical ontologies as well as referencing ontologies such:

• the Protein Ontology (PR) for uniquely expressed biomarkers

• Gene Ontology (GO) for the biological processes a cell type participates in.

Things we describe in Biology – Pathways

• Reactome is a database of pathways

• Has export to BioPax ontology to describe pathway elements

• Connects many biological concepts including nucleic acids, genes, disease and GO terms

OBO Foundry• OBO = Open Biomedical Ontology

• The OBO Foundry seeks to organise human expertly curated ontologies in biomedicine

• Provides a set of principles for best practice

• Six OBO Foundry ontologies

• OBO library much bigger and there are many Foundry candidate ontologies

• Intrinsically, biology is interconnected yet many ontologies are not formally linked

• Ontology development is expensive – reducing overlap and improving collaboration would decrease this

• Modularity of domains would increase reusability

Let 100 flowers bloom vs Centralised collaboration• 100 flowers bloom:

• Competition driven

• Application and data driven (often to local use cases)

• Requires no commitment to upper ontology framework

• Mapping between efforts can be costly (potentially exponential)

• Duplication of effort

• Centralised collaboration:

• Encourages collaboration and openness

• Aim to produce consensus model of domain knowledge

• Reducing overlap reduces duplicated effort

• Interoperability part of methodology

• Requires upper ontology commitment

• Development by committee can be inhibiting