This page contains a list of FAQ and answers that came up during the workshop in discussions and as part of a wider question session with Robert Stevens.
What is an ontology?
An ontology is a model of common entities. It provides a formal vocabulary to describe entitites (e.g. genes) and the concepts that surround them (e.g. gen function or process).
Why should I bother with ontologies?
Data are useless unless we describe what they are. We also need to agree on the concepts that describe the data, and the labels used for these concepts. Ontologies are a logical approach to standardising the labels we use.
What is the difference between an ontology and a thesaurus?
An ontology describes both the entities and the relationships between them. A thesaurus does not make the distinction between a concrete thing (e.g. an enzyme) and a process (e.g the catalysis of a reaction).
What can ontologies describe?
Ontologies can describe any or all of the following:
- Entities themselves - genes, proteins, enzymes, metabolites
- Investigations that produced the entities - sequencing, mass spectroscopy, functional assays
- Information on the history and origins of the entities and their descriptions - raw data and provenance
What kinds of information can ontologies capture?
Ontologies can capture all of the following information types:
- Process - Gene Ontology: Biological process
- Information - Information Artifact Ontology
- Material - Chemical Entities of Biological Interest (ChEBI); NCBI Taxonomy; Gene Ontology: Cellular component
- Property - Gene Ontology: Molecular function; Phenotypic Quality: PaTO, Human Disease Ontology
- Site - Gazetteer Ontology
Can I do what ever I want with an ontology?
Most ontologies are open and free to use. There might be restrictions on certain usage, or for re-distribution of the ontology.
When is it better to knock together my own ontology vs reuse something that exists already?
It is generally better to reuse an exisiting ontology if you can, rather than starting from scratch. This may mean the ontology may not fully capture or describe what you wish it to. Alternatively, developing a new ontology allows you the freedom to define your own terms and structure. Terms can be crosslinked to exisiting ontologies if required. This approach requires time, but it may be faster than waiting for the community to develop one. Another factor to consider is the cost of licensing a proprietary ontology.
Will multiple ontologies be consolidated in the future?
Whilst there are many different ontologies, they all have reasons why they were developed. In certain cases an ontology may be used and further developed for different reasons. For example, ChEBI exists as a free public ontology, while industry makes use of many other proprietary chemical ontologies with their own descriptors. It is unlikely that industry will give up using their own ontologies.
Is it possible to link ontologies together?
This is an ongoing research problem. Sometimes it is possible to find common ground between two different ontologies. Building an ontology as part of a use case is often easier than leveraging two existing ontologies together.
Can an ontology be simplified?
Terms can be removed from the ontology and relationships between entities can also be removed.
What happens when you delete terms from an ontology?
OBO has a rules covering the deletion of terms from an ontology. A term should be obsoleted and marked as not used, rather than being deleted from the ontology.
How should you prioritise annotation?
When creating an ontology, the best approach should be to annotate terms that are easy to define and will form the core of the ontology - the 'low-hanging fruit'. It is also sensible to prioritize annotations that are in most demand by potential users - surveys or prior research might give an indication of these terms. More information can be found in James Malone's slides.
How can expert information be gathered from those without ontology experience?
Tools exist that provide simple ways to collect and define ontology terms. These don't require the data provider to have a detailed knowledge or working understanding of ontologies. More information can be found in Simon Jupp's slides and on the owlpopulous webpage.