CELL TYPE ONTOLOGY (CTO) NORMALISATION EXPERIMENT
Intro
In this experiment we are recreating the Cell Type Ontology using a Normalisation pattern. In a normalised ontology the is-a (subsumption) structure is generated by the reasoner rather than by hand. This has various advantages:
- Consistent maintenance. The reasoner infers all the subsumptions, so there are no missing is_a links, if the axioms are correct.
- Adding new classes is easier, as the ontologist need only to follow the object properties and add fillers, without having to explore the whole structure. Also, that means that richer querying can be performed, because there are many more object properties in the ontology.
This experiment is not intended as a substitution of CTO, it is just an experiment on ontology engineering, a demonstration of a different method for building ontologies using OWL and a reasoner.
In order to perform the normalisation, an ontogenesis meeting (meeting 5) was celebrated with the following participants: Mikel Egana, Simon Jupp, Robert Stevens, Helen Parkinson, Phil Lord, James Malone, David Shotton. In such meeting a normalisation scheme was decided (see the notes) consisting on a set of properties that can be used to define cells (e.g. ploidy, function, etc.). The "leaf" cells of CTO were divided in groups, and each participant volunteered to add the fillers for each property of the scheme to the cells of the chosen group (e.g. secretory cells), following a guideline.
Pipeline
The pipeline adds the cells defined by each participant (in an spreadsheet) to an empty OWL ontology, in order to recreate CTO from scratch via reasoning. The pipeline adds many more axioms (e.g. develops_from closure) and ontologies (FMA, GO, PATO, RO) to obtain meaningfull results via reasoning.
The implementation of the pipeline is a very complex and dirty mess of onto-perl and OPPL scripts, but if you want to have a look into it send me an email (mikel.eganaaranguren@cs.man.ac.uk).
Remaining cells
We have done most of the leaf cells (Simon is finnishing the groups). Now we are doing intermediate cells, to recreate via reasoning, as much as possible, the original CTO structure. Helen is working in the hierarchy under contractile cells and the one under Barrier cells, and Mikel is working in the hierarchy under secretory cells.
If you want to add intermediate cells first drop a line to mikel (mikel.eganaaranguren@cs.man.ac.uk) so we don't do redundant work. Then, fill an excell file following the same template as the leaf cells and the guideline in http://www.gong.manchester.ac.uk/CTO_guideline.txt.
Resulting ontologies
The resulting ontologies are updated very frequently so please stay tuned for the latest. The main result (cto_norm.owl) imports the other ontologies so Protege should resolve them when opening. Download.