GO Sensu Term and Taxonomy (Apr 2007)
Download as PPT, PDF1 like341 views
The document discusses two related projects regarding the use of taxonomy in the Gene Ontology (GO): 1. Eliminating references to taxon from GO term names and definitions. There is consensus that this is desirable, and the remaining terms need to be updated. 2. Determining how or if GO should associate terms with taxonomic groups. Two positions are presented: either eliminate all taxon associations, or formalize the associations in terms like "relevant_for" and use them for subsets/slims and annotation. Key open questions include how to handle taxon information currently in definitions, the purpose and risks of subsets/slims, and ensuring consistent treatment of taxonomy across OBO ontologies.
GO Sensu Term and Taxonomy (Apr 2007)
- 1. the end of sensu & what to do about taxonomy Ontology content meeting GO Apr 2007
- 2. Two distinct but related projects Part 1 Elimination of taxon as a differentium Desirability Feasibility Timeline Part 2 Linking OBO terms to taxons Framework Should GO do this What to do about subsets?
- 3. Taxon is not a good means of discriminating Before: Taxon used in name and definition cell wall (sensu Magnoliophyta) cell wall (sensu Bacteria) After (ie now): Taxon eliminated from name and definition cellulose and pectin-containing cell wall peptidoglycan-based cell wall (vestigial trace of taxonomic info in definitional gloss ) (old names retained as synonyms) Let’s call these “ post-sensu terms ” for the purpose of this discussion A more or less rigid stucture enclosing the protoplast of a cell and composed of cellulose and pectin and other organic and inorganic substances . As in, but not restricted to, the flowering plants (Magnoliophyta, ncbi_taxonomy_id:3398 )
- 4. Consensus check Check: we all agree this is a good thing to do in principle? Check: do we think this is possible to do for all current sensu terms? 153 remaining to de-sensu-itize Put up difficult cases on sensu wiki page E.g. gametogenesis in moss and fern Do we have any need for sensu as purely lexical disambiguator Eg sensu-community; neurogenesis Question: what (if anything) do we do about the taxonomic information? Currently being relegated to definition gloss .
- 5. GO currently has some kind of association to taxons In the Ontology Subsets Goslim_yeast Goslim_plant Gosubset_prok Imprecisely defined association Sensu (now definitional gloss) Semi-precisely defined (“as in, but not limited to…”) In the Annotations Via gene products Precisely defined (more or less)
- 6. If GO terms are to have any kind of association to taxons then the meaning of that association should be clear Meaning should be clear for annotations Meaning is less clear for subsets and post-sensu terms E.g. 417 unpropagated subsets Intentional?? People are already using this information - and possibly in the wrong way
- 7. position #1: No to taxa If GO terms are to have any kind of association to taxons then the meaning of that association should be clear Therefore we should eliminate all possible GO to taxon links Phase out taxon slims/subsets Eliminate taxonomic info from definitional gloss of post-sensu terms Annotations are the only true source Problem solved!
- 8. position #2 : Yes to some taxon links If GO terms are to have any kind of association to taxons then the meaning of that association should be clear Some kind of taxonomic association is useful to some people With caveats Therefore we should make this association clear We need to define what it means to link a taxon to a term Relevant_for Valid_for Specific_to Canonically_found_in … . We should export our methods to other OBO ontologies
- 9. Relevance (applicability) Term relevant_for Taxon: Instances of Term found in some species of that Taxon Current semantics for taxon-subsets? Current semantics of post-sensu taxon info? Examples: peptidoglycan-based cell wall relevant_for Bacteria Suckling behaviour relevant_for Mammalia This is a weak association Hatching relevant_for Mammalia Because at least one Mammalian species has members that hatch We can understate cell wall relevant_for Bacteria (this is true but under-specified)
- 10. Perhaps this is all we need for GO But let’s look at some more
- 11. specificity Term specific_to Taxon: Instances of Term found only in Taxon Examples: Apoplast specific_to Viridiplantae Suckling behaviour specific_to Mammalia We can err on the conservative side: Sucking behaviour specific_to Metazoa Counter-examples Maternal behaviour NOT specific_to Mammalia
- 12. validity Term valid_for Taxon: Instances of Term are found in some organism of all sub-taxons of Taxon Examples Suckling behaviour valid_for Mammalia (all Mammalian species suckle) Trivially true: Biological process valid_for Viridiplantae
- 13. As found in, but not restricted to We are currently retaining this info in the gloss of the post-sensu terms E.g. As in, but not restricted to, the flowering plants (Magnoliophyta, ncbi_taxonomy_id:3398 ) Semantics same as relevant_for Perhaps stronger - we can say the Term is exemplified in the Taxon Intended primarily for humans?
- 14. Difficulties Taxonomies are subject to revision Biology keeps turning up surprises But this is the same as any other part of GO?
- 15. What should GO do? Continue with eliminating taxa from names and definitions What is our policy for taxon-subsets? Eliminate? At least formalize in terms of relevant_for? What should we do about the taxonomic info consigned to definitional gloss? Get rid of it? Formalize it in terms of relevant_for, allowing it to be used for dynamic slim-generation? Should we annotate other kinds of GO-taxon links? Validity, specificity?
- 16. Open questions What are subsets/slims for anyway? Are they dangerous? Eg annotations to “leaf sensescence” would be mapped to biological_process if we use goslim_plant Are there better ways of making them? Combination of annotations, specificity and curated GO term <-> taxon associations
- 17. Other OBO ontologies Many OBO ontologies are using sensu in the old, bad way Mammalian Phenotype Plant anatomy Other OBO ontologies are more entwined with phylogeny and homology E.g. ZFIN CToL fish anatomy ontology We want consistent usage across OBO
- 18. Summary Elimination of taxons from names and definition Are we agreed? What, if anything, do we do with the taxon information in the definitional gloss? What is our policy on subsets/slims? GO should lead the way for (or at least be consistent with) OBO
Editor's Notes
- #2: And I’m going to be as dry and impersonal as possible. This is a purely technical talk. Mark gave his personal views and admitted to being confused I’m not sleeping either. We want very much to work together. Ontology development is about collaboration.