The Solution: Hymenoptera Anatomy Ontology
From HAO Wiki
A modest proposal
At the 2006 International Conference of Hymenopterists (ICH) we (Deans & Ronquist 2006) proposed a group effort to consolidate the terminology we use for Hymenoptera anatomy, and the Hymenoptera Ontology (hereafter HAO) project was initiated. The HO remains an official project of the International Society of Hymenopterists (ISH), which guides the ontology's development with advice from a seven-member committee (PI Deans serves as the chair). Our initial objectives for this effort were to 1) converge on a common vocabulary to allow for easier comparisons across a broad array of Hymenoptera and organize this vocabulary into a directed acyclic graph (DAG), 2) partner with Morphbank to facilitate more efficient image searching using the DAG, 3) create an online glossary and browsable atlas of hymenopteran anatomy to serve as the "official" (i.e., member voted) resource for ISH and anyone interested in hymenopteran anatomy.
Early Development of the HAO
Our initial strategy for ontology development (i.e., prior to the proposal at the ICH) was for one person (PI Deans) to assemble the HAO using DAG-edit (now OBO-edit; http://oboedit.org) and a long list of 600+ terms manually extracted from classic treatises on Hymenoptera (and other insect taxa) comparative morphology (Snodgrass 1935, Duncan 1939, Gibson 1985, Johnson 1988, Vilhelmsen 1996, Schulmeister 2003, Mikó et al. 2007). Terms were keystroked into DAG-Edit, textually defined, and related using part_of, is_a, and synonym. We also implemented an obsolete_synonym relationship (semantically equal to synonym) for terms that created confusion (multiple homonyms, e.g.) and should be abandoned; Schulmeister (2003; pg. 156) provides an excellent example of such a set of terms (reproduced here as Table 2).
During the early stages of ontology development, however, we quickly determined that a simplified, centralized, and more accessible tool was required to effectively encourage input from the hymenopterist community— i.e., that the HO cannot be developed in isolation. Deans and Ronquist established a partnership with M. J. Yoder to develop and gradually expand a prototype ontology builder in mx (http://purl.oclc.org/NET/mx-database). This new software component, the Hymenoptera Ontology Database (HAO DB), integrated existing tagging and figuring functionality for data objects in mx to allow for commenting and suggestions for alternative definitions. A query interface to Morphbank was added to mx, which allowed references to be made to images stored on Morphbank servers.
Access was requested and provided to more than 40 hymenopterists, and several attempts to recruit content were made through short talks at annual meetings. Only ten hymenopterists have provided data so far, with the vast majority of terms, references, relationships, comments, and images derived from the three authors of this proposal. The lack of input, however, does not accurately reflect the level of enthusiasm bubbling through the community. The limited development stems mainly from a lack of understanding: What tools are being developed? How do I contribute? How can my lab exploit this resource? The HAO project requires an infusion of resources, knowledge, and funding to mature and develop a more broadly usable toolset and, more importantly, to educate its potential users.
First Tools to Employ the HAO
Following several months of contributions to the HO DB we decided to expose the existing data as a "Hymenoptera Glossary". To further illustrate the potential utility of the HO DB Yoder built a text mark-up tool in mx that links terms in blocks of text to records in the Hymenoptera Glossary (Yoder, 2007). The markup tool can be used for any text within mx, including phylogenetic character descriptions. The utility of the markup functionality was made public with a simple "proofing" tool. This function allows users to paste in a block of text to be marked-up against terms in the database and serves as a simple and straightforward means to check whether, for example, the terms in one’s description are also in the database. More recently, Yoder built an XML-based search interface to Morphbank. The interface provides a simple drag/drop mechanism for linking images to objects in mx (most importantly "terms" in the context of this proposal). With text annotations already available within the figuring functionality Yoder also added the ability to graphically annotate images, functionality that allows for the precise delimitation of areas of interest in the context of a given definition or term.
The HO has grown to include 1,925 terms with >2,000 relationships. The development process exposed interesting details about how we use our domain language, which otherwise would have remained hidden. Emerging homonyms (e.g., speculum, pedicel, gaster, face, stigma, disc, metapleural triangle, anellus) refer to often extraordinarily disparate anatomical entities. Chaotic character systems (e.g., propodeal ridges, pronotal ridges, glands, occipital carinae, cuticular patches, male and female genitalia, thoracic musculature) beg clarification, especially in that these systems are often rich sources of diagnostic and phylogenetic characters. In many instances the same term is used to describe homologous structures, e.g., 'margin of the clypeus' in Formicidae and Proctotrupomorpha (Ronquist & Nieves- Aldrey 2001; Alpert 2007) or 'margin of the pronotum' in Proctotrupomorpha and Ichneumonoidea (Perrichot et al. 2008; Gauld 1976). These terms are qualified inconsistently, however, depending on the perceived orientation of the structure: ventral margin of the clypeus in gall wasps, which have a hypognathous head orientation, vs. anterior margin of the clypeus in ants, which have a prognathous head orientation.
Also revealing are results from new species descriptions where the Hymenoptera Glossary proofing tool was iteratively employed for feedback. Though we have only a small sample (n=4), taxon descriptions were eventually enriched by a 56% increase in obviously semantic statements about anatomical phenotypes (based on the user’s interpretation that initial descriptions were inadequate with respect to the contents of the HO). We are positioned to take the next step in the development of this resource and expect that the HO will facilitate a similar revision of future species descriptions.
Table 2. Obsolete synonyms of 'tarsal plantula' extracted from Schulmeister (2003).
|tarsal pulvillus obsolete_synonym tarsal plantula||'pulvillus' also used for 1) the paired (basi)pulvilli of the pretarsus in other insects, situated below the claws, and 2) to indicate the arolium, a pad between the pretarsal claws|
|plantar lobe obsolete_synonym tarsal plantula||'planta' is used as a name for 1) the ventral face of the tarsomeres, and 2) a sclerite on the arolium and the basal tarsomere of the hind leg of worker bees|
|euplantula obsolete_synonym tarsal plantula||'euplantula' used for different structure in other Neoptera|