Using Phenotype Ontologies in GVF
This page describes developing a set of best practices when using phenotype ontologies in GVF.
The first draft of these recommendations written by Chris Mungall A second round of this draft was made during a Phenotype RCN collaboration meeting (September 16-18, 2013 - with Chris Mungall, Melissa Haendel, Matthew Brush, Mike Bada, Bret Heale and Karen Eilbeck). The document has been split into Human Phenotype annotation and Non-human phenotype annotation.
Using IDs vs Term Names vs Comments
GVF allows you to use either IDs or label strings to indicate the phenotype. I recommend using the ID at all times.
If the ontology term is not granular enough, use the comment field to enter a free text description
Which Phenotype Ontology to use?
If we take "phenotype" to be inclusive of traits, diseases, pathological features, etc then there are a number of choices. The main ontologies are listed on the following 2 pages, together with examples.
Compositional Descriptions of Phenotypes
Many phenotyping groups describe phenotypes using post-composition - the simple case is where a general characteristic (e.g. "hyoplastic", from PATO) is combined with an anatomical term (e.g. "kidney", from FMA). This is in contrast to ontologies such as the MP, which pre-compose phenotype descriptions, with a single ID for each description. For more details see:
- Mungall, C. J., Gkoutos, G. V., Smith, C. L., Haendel, M. A., Lewis, S. E., and Ashburner, M. (2010). Integrating phenotype ontologies across multiple species. Genome Biology 11, R2 
At this time GVF does not support post-composition. Note that post-composed descriptions can always be translated into a phenotype ontology outside of GVF. However, some groups might still feel the need to use post-composition within GVF. The developers are open to extending it in this direction in the future - contact them on the mail list below.
One way in which compositional descriptions are useful is for describing measurements
Quantitative description of phenotypes
Pre-composed phenotype ontologies typically use qualitative descriptions, such as "increased width of face". Sometimes it's useful to record actual measurements - this could be a tuple of entity - characteristic - unit - value.
Currently GVF doesn't support this, but support could be added in the future.
Hancock, J. M., Mallon, A.-M., Beck, T., Gkoutos, G. V., Mungall, C., and Schofield, P. N. (2009). Mouse, man, and meaning: bridging the semantics of mouse phenotype and human disease Mammalian Genome 20, 457-461 
Including Phenotype Annotations outside GVF
Of course, it is always possible to include phenotype annotations outside GVF. This has the advantage of allowing for a more expressive representation, including evidence, assay details, temporal information, measurements, complex compositional descriptions etc.
However, as yet there is no one single format that suits all requirements. There is a growing body of best practice on how to record these using OWL - contact the obo-phenotype mail list below if you are interested.
In the absence of a single recommended supplementary phenotype information file, we recommend including as much phenotype information as is appropriate directly in the GVF file using one of the ontologies above.
You can ask questions about GVF on the song-devel mail list (song-devel AT lists DOT sourceforge DOT net)
Questions about phenotype ontologies can be asked on the obo-phenotype mail list (obo-phenotype AT lists DOT sourceforge DOT net)