Ontologies for Ethology

Peter E. Midford

Introduction

Welcome to the ontologies for ethology page, documenting a project to develop knowledge representation, including ontologies, as tools for describing and analyzing animal behavior. The project is currently developing a set of ontologies for courtship behavior in a clade of Habronattus jumping spiders from raw video of behavior sequences. Ultimately, a small library of such knowledge bases will serve as sample data to develop methods for comparative studies and (secondarily) other varieties of linking between knowledge bases across behavior studies. The plan is to eventually provide these comparative tools into the Mesquite tool.

The purpose of this project is not to propose a particular upper-level or coding for ontologies, but to demonstrate ontology construction as a general technique for coding ethograms and other descriptions of behavior into machine understandable forms. As such, it is intended to encompass, rather than serve as an alternative to, previous attempts ^1,2 at formalizing the description of behavior.

The goals of this project differ from efforts such as those of the Gene Ontology Consortium. This project is more focused on representing and coding descriptions, rather than using ontologies as a semantic level of metadata for linking entities in existing databases. Part of this reflects the small number of data sources in this project. A plan to work on the integration across sources of description is coming in the near future.

New

I presented a talk on comparative methods for ontologies at the 2007 Animal Behavior Society meeting. I have uploaded the Powerpoint both as a PowerPoint file and as an HTML slide show.

Looking for information about OwlWatcher or EOLite? Try here

Sample Ontologies

This site currently includes two ontologies developed as part of this project.

Loggerhead Nesting

An ontology for Loggerhead sea turtle (Caretta caretta) nesting behavior, based on the published ethogram of Hailman and Elowson³ is available in three forms:

An abbreviated ontology (just concept classes) of the actions and states in the ontology. This is rendered in HTML with one page for each concept defined in the ontology. This is the recommended place for people who know about behavior, but not about ontologies, to start.
A complete dump of the ontology and knowledge base as a HTML tree - one page for each concept and instance.
Protege-2000 source (CLIPS) files - you will need the Protégé-2000 editor to use these files. Brief directions for loading them are here.

History

This Protégé version is a revised version of the Loom based ontology presented in posters at the Evolution 2001 and Animal Behavior 2001 meetings. Further updates to this ontology may be posted from time to time.

Substantial updates to the Loggerhead ontology were made during December 2001 and September 2002.

Habronattus Courtship

A draft version of an ontology for Habronattus californicus courtship is available in three formats:

An abbreviated ontology (just concept classes) of the actions and states in the ontology. This is rendered in HTML with one page for each concept defined in the ontology. This is the recommended place for people who know about behavior, but not about ontologies, to start.
A complete dump of the ontology and knowledge base as a HTML tree - one page for each concept and instance.
Protege-2000 source (CLIPS) files - you will need the Protégé-2000 editor to use these files.

History

This ontology is part of an ongoing comparative study of courtship behavior in a clade of eight Habronattus spiders, which include H. californicus and H. clypeatus. The later will be the next ontology to be posted. This ontology was described at the 2002 meetings of the American Arachnology Society, the Animal Behavior Society and at Measuring Behavior 2002.

Miscellaneous Introductory Material

How to read the HTML rendered ontologies

The generated rendition of the ontology begins with an outline of the concepts defined in the ontology, organized by superclass/subclass relations. Within the page for each concept, there is a textual definition, followed by a list of slots for that concept. Each slot specifies a relation that may (or must) exist between instances of the current class and other objects described in the ontology or its associated knowledge base. Each slot is specified by a name, a description, and several other properties (also called facets of the slot). The "Type" property specifies the "data type" of the object that fills it - usually instance, less frequently class or number. The "Allowed values" property specifies the classes that can fill a slot of "Type" instance. This is a more "implementation transparent" notion of type. Finally, the cardinality specifies the number of things that the slot can refer to: the token 0:* means the slot can refer to any number of things.

Other material available on this site

An overview (PDF file) of ontologies I made available at the Ethoinformatics workshop held in April 2002 in Bloomington IN.
A short list of notes describing the design and construction process for the ontologies may be found here.
A longer version of the short note I published in Bioinformatics is available here. The original (Midford, P. E., 2004. Ontologies for Behavior. Bioinformatics 20:3700-3701.) is available from Bioinformatics here.

On-Going Projects and Future Work

Loggerhead Nesting Although the coverage of the behavioral terms is complete, this ontology is useful as a test bed for developing software tools and advanced representation ideas, such as prototypes. Prototypes are particularly important in published ontologies since the descriptions represent a summary of multiple events.
Habronattus courtship The process of coding courtship videos is ongoing, producing conventional narratives as well as ontology representations. This dataset will be used for development of comparative tools based on ontologies.
Comparative Methods The primary goal of the work described here is to develop comparative methods for ontology-based behavior data. Comparative methods for ontologies can be approached in three ways:
- Extract information from the ontologies into a conventional character matrix.
- Align (assert homologies between) terms in the ontologies for a set of species and count changes. This method is analogous to conventional parsimony methods.
- Ontology based models of evolutionary change. Here ontologies provide a language for describing models of evolutionary change (e.g., actions and the sequences that contain them) and also constraints on transitions or states within the model. For example, courtship sequences should contain a copulation act or the functional equivalent. A model should prevent the copulation act from being deleted, though it might allow other acts in the sequence to be deleted. This constraint would use terms from the ontology to define just what couldn't be deleted. A smarter constrain would allow the copulation act to be replaced by something functionally equivalent, such as dropping a spermatophore and causing the female to accept it.
Slides from a talk that describes these methods is here.
Other Ontology Tools I have shifted away from frame-based Protege at this point. I am currently working with OWL and OBO format files. I am developing a tool called OwlWatcher that combines a simple ontology editor with a video clip player. It currently works with OWL format ontologies, but I am building a plugin that will allow processing OBO format files as well.

Acknowledgements

I wish to thank Jack Hailman for permission to tear up his ethogram, Wayne Maddison for time and opportunities to work with the Mesquite project. They also provided the images of their study animals found on this page. These ontologies were constructed while I was funded by a National Science Foundation Bioinformatics Postdoctoral Fellowship. Any opinions, findings, and conclusions or recommendations expressed herein are those of the author and do not necessarily reflect the views of the National Science Foundation. More recent support has come from:

Cipres Project and my postdoctoral positions with Wayne Maddison and Mark Holder.
Part-time support from the Phenomap project.

References

1. Golani, I. 1976. Homeostatic motor processes in mammalian interactions: a choreography of display. In: Perspectives in Ethology vol 2. (Ed. by P.P.G. Bateson & P. H. Klopfer), pp. 69-134. New York: Plenum.

2. Schleidt, W. M., Yakalis G., Donnelly M. & McGarry, J. 1984. A proposal for a standard ethogram, exemplified by an ethogram of the bluebreasted quail (Coturnix chinensis). Zeitschreift für Tierpsychologie 64:193-220.

3. Hailman, J. P. & Elowson, A. M. 1992. Ethogram of the nesting female loggerhead (Caretta caretta). Herpetologica 48:1-30.

All ontologies on this site and content of this page (except where otherwise acknowledged) are copyright 2001-2007 Peter E. Midford (peteremidford@yahoo.com).
Last update of this page: 18 October 2007