One of the most challenging problems now facing the model organism databases is the formal description of phenotypic data. While some databases, for example those for mouse (Mus musculus) 1, corn (Zea mays) 2 and fruit fly (Drosophila melanogaster) 3, include a rich heritage of data describing the phenotypes of mutants, and some progress is being made to bring these data into a well structured computable representation 345, the annotation of these phenotypes is hampered by a lack of structured information describing a variety of other biological objects, including cell types. A structured vocabulary of cell types is also required by databases for the description of other biological objects, such as gene-expression data. In addition, using the same concepts for the description of these data in all of these databases would facilitate interoperability among them.

To address these needs, we have developed an ontology that describes the cell types of the major model organisms, both animal and plant. Its use will allow a biologist to query a single database with such questions as: list all of the cell types in mouse that express the Notch gene and all of the cell types in Drosophila and Caenorhabditis elegans that express the closest homolog of this gene; list all of the genes in mouse, rat, human and zebrafish that are expressed in the cell type Schwann_cell; CL:0000218; list all of the genes in D. melanogaster and C. elegans that have a mutant phenotype in the cell types that develop from the cell type myoblast; CL:0000056. The use of the cell ontology will thereby promote the de facto integration of data from diverse databases.

Since the development of the Gene Ontology (GO) for the annotation of attributes of gene products 6, many ontologies have been developed in the model organism informatics community. Several of these are available, in a choice of common formats, from the Open Biological Ontologies (OBO) site 7. They include comprehensive developmental and anatomical ontologies for many model organisms (for example, mouse, Drosophila, Arabidopsis thaliana and C. elegans), and ontologies for mouse pathology and human disease.

There are several other ontologies that include cell types such as Systematized Nomenclature of Medicine (SNOMED) 8, the Foundational Model of Anatomy (FMA) 9, the anatomy ontologies used in model organism databases at the OBO site 7, vocabularies used by the resources that hold cell lines such as the American Type Cell Collection (ATCC) or the European Collection of Cell Cultures (ECACC) 1011, and others 1213. Our approach for handling cell types differs from that adopted by these resources. First, SNOMED, FMA and the species-specific anatomy ontologies explicitly assume that the cell types they include are associated with one particular organism. Their identifiers cannot therefore be used to annotate cell types from other organisms, even if these cell types are essentially identical to those in the organism-specific ontologies. Second, these resources, together with those that hold cell lines (for example, ECACC and ATCC), tend to define cell types as constituents of tissues rather than provide phenotypic information about their attributes - the knowledge that they encapsulate is severely limited. Third, some ontologies do not have publicly available identifiers for each term; hence they cannot be used for general annotation 1011. The Plant Ontology 14 provides a cell type node that shares some of the organizing principles of our cell ontology, but it is limited to those cell types found in plants. For all these reasons, we set out to produce an organism-independent ontology of cell types based on their properties (such as functional, histological and lineage classes) and report here the availability on the Open Biological Ontologies site 7 of this ontology, which incorporates the cell types possessed by a broad range of phyla and is defined by a rich set of criteria.

Ontologies in bioinformatics are intended to capture and formalize a domain of knowledge, and the ontology reported here attempts to do this within the domain of cell types. It is designed to be useful in the sense that a researcher should be able to find, in a rapid and intuitive way, any cell type in any of the major model organisms and, having found it, learn a considerable amount about that cell type and its relationships to other biological objects.

A core feature of the ontology, and one that differentiates it from other resources that contain cell types such as SNOMED and the FMA 89, and the Drosophila and Arabidopsis ontologies 317, is that the cell ontology explicitly sets out to include cell types from all the major model organisms within a common framework. In addition, it also seeks to incorporate a great deal of phenotypic information about these cell types and is thus far more comprehensive in its cellular detail than these other resources. The intention is that the new cell-type ontology should provide organism-independent knowledge as well as cell-type unique identifiers (ID) that can be incorporated into any database holding cell-type-associated knowledge. The formalized structure of the ontology, together with its set of unique IDs, will allow curators to incorporate cell-type data into their databases, integrate the data with the knowledge encapsulated in the ontology, and use the IDs to interoperate with other databases. While we expect such bioinformatics applications to be its immediate use, we hope that, in the longer term, all biologists will find the ontology useful.

The expected short-term use of the ontology will thus be in cataloguing phenotypes and gene expression patterns. Indeed, it is quite surprising that those who work with model organisms still lack the bioinformatics resources needed to catalogue, archive and access the details of the phenotypes emerging from mutant screens and natural variations. A robust representation of normal and mutant phenotypes in all of the model organisms will require ontologies for a wide range of macroscopic properties (pathology, anatomy, abnormal quantifiers, and so on) and we view the cell ontology as a component of this programme that should be useful in cataloguing phenotypes (and other attributes) associated with cell types.

In the long term we expect that molecular biology and biological databases will move beyond being gene-centric and that biological mechanisms will be studied at a more integrated level. Cells are the biological units with which tissues and organs and organ systems are built. A rich and explicit description of cell types across phyla that are adapted by biological databases will help facilitate this transition.

Finally, it should be pointed out that, like many such resources, this ontology is not complete: although it contains all the common cell types, there will certainly be some that have been omitted. Most importantly, although many of the cell types are fully described by function, morphology, organism, and so on, others are inadequately described and more relationships need to be made. A particular weakness is the fact that the category identified as experimentally_modified_cell has yet to be populated, and doing this will involve consideration of the various cell lines held in the major collections. As with other community resources, community input is essential for the development and maintenance of the Cell Ontology; biologists with comments and additions are therefore welcome to contribute to the ontology and should contact the curator ashburner@ebi.ac.uk.

The ontology includes the major cell types from the major model organisms (for example, human, mouse, Drosophila, Caenorhabditis, zebrafish, Dictyostelium discoideum, Arabidopsis, fungi and prokaryotes). These cell types have been collated from our own knowledge, from major textbooks (for example 202122), from the embryo and anatomy ontologies available on the OBO site 7, and from colleagues (who are thanked in the acknowledgements). The ontology currently holds some 680 cell types, together with their synonyms and, in most cases, text definitions.

The ontology was constructed using the open source Java tool OBO-Edit (previously known as DAG-Edit) 23, which is convenient for building ontologies that are consistent with the GO formalism. The resulting ontology is available in both the GO 'flat-file' format 24 and the newly defined 'OBO format' 25, and can easily be viewed using the OBO-Edit or the COBrA open source Java tool 26.

The Cell Ontology is available from the OBO site 19. Following the cell.obo link will take the user to a page in which the current version of the Ontology, and archived older versions, can be viewed (view) or downloaded (download). Differences between the current and previous version can be seen by following the Diff to link.