Plants from a handful of species provide the primary source of food for all people, yet this source is vulnerable to multiple stressors, such as disease, drought, and nutrient deficiency. With rapid population growth and climate uncertainty, the need to produce crops that can tolerate or resist plant stressors is more crucial than ever. Traditional plant breeding methods may not be sufficient to overcome this chal-lenge, and methods such as high-‐throughput sequencing and automat-ed scoring of phenotypes can provide significant new insights. Ontolo-gies are essential tools for accessing and analysing the large quantities of data that come with these newer methods. As part of a larger project to develop ontologies that describe plant phenotypes and stresses, we are developing a plant disease extension of the Infectious Disease On-tology (IDOPlant). The IDOPlant is envisioned as a reference ontology designed to cover any plant infectious disease. In addition to novel terms for infectious diseases, IDOPlant includes terms imported from other ontologies that describe plants, pathogens, and vectors, the geo-graphic location and ecology of diseases and hosts, and molecular func-tions and interactions of hosts and pathogens. To encompass this range of data, we are suggesting in-‐house ontology development comple-mented with reuse of terms from orthogonal ontologies developed as part of the Open Biomedical Ontologies (OBO) Foundry. The study of plant diseases provides an example of how an ontological framework can be used to model complex biological phenomena such as plant disease, and how plant infectious diseases differ from, and are similar to, infectious diseases in other organism.
Plants are the primary food source on which almost every
other organism on earth depends, either directly or
indirectly, and six plant species – wheat, rice, corn, potato, sweet
potato, and cassava – provide 80% of calories consumed by
humans worldwide
Many challenges in plant pathology (the study of plant
diseases) can potentially be met through advances in
methods such as high-throughput sequencing and automated
scoring of phenotypes
A plant disease is traditionally defined as a deviation
from normal physiological functioning that is harmful to a
plant
IDOPlant will integrate and interoperate with member and candidate ontologies of the Open Biomedical Ontologies (OBO) Foundry (Table 1), such as: the Plant Ontology (PO; describes the plant structures and the development stages at which infections happen or signs of disease are observed), the Plant Trait Ontology (TO; describes phenotypes or entities that are evaluated in plants, such as leaf color or grain yield), and the Gene Ontology (GO; describes Ontology Name
ID
PO
GO
the molecular functions of interacting genes from host and
pathogen as well as biological processes involving either
host, pathogen, or both). The multi-organism process branch
of the GO, developed as part of the PAMGO project
The study of plant diseases provides an excellent
example of how the framework of the OBO Foundry
Throughout this paper, words in italics are ontology terms,
e.g., pathogen. If the source ontology is not evident from
the context, then we prefix with the ontology ID, as in:
IDO:pathogen. The IDOPlant and the Plant Phenotype and
Stress Ontology are being constructed in web ontology
language (OWL), using the Protégé 4.1 software
(http://protege.stanford.edu) The annotation standard
format will follow the GO and PO model with the GAF2.0
format
We began by reviewing whether the terms in the IDO were adequately structured for describing plant infectious diseases, including discussion with the developers of IDO and the Ontology for General Medical Science (OGMS; http://code.google.com/p/ogms/). Next, following the strategy used in other IDO extensions, we created terms for the IDOPlant, such as plant infectious disease, specific to the needs of plant pathology. More specific terms, such as rice bacterial blight disease were added as an example of how to model a specific disease and to provide terms to be used in annotating existing gene expression data available through Gramene (http://www.gramene.org). Textual definitions and relationships among terms are drawn from plant pathology textbooks or journal articles, and are reviewed by plant disease experts.
Logical definitions for IDOPlant terms are being
constructed in OWL. Many of the terms needed for logical
definitions already exist in other ontologies. To access these
terms, we could import entire ontologies into the IDOPlant,
but this would result in many unnecessary terms and may
cause problems if the resultant ontology is too large.
Importing a selected subset of terms creates problems as well. If
we import individual terms from external ontologies, then
we lose the ontology structure they are associated with and
the reasoning power that comes with it. If we import
selected terms through the MIREOT process
To cope with these issues, we use a multi-pronged
strategy that includes directly importing some terms and
building bridge files to link to external ontologies.
• Terms specific to plant diseases are added to the
IDOPlant and assigned unique IDOPlant IDs, e.g.,
IDOPlant:#######.
• Terms falling near the bottom of the IDOPlant hierarchy
that are drawn from ontologies from which only a few
terms are needed are imported as single terms, using the
original ontology ID. When appropriate, the MIREOT
method is used.
• Content treated in ontologies from which many terms are
needed are accessed by simultaneously loading multiple
ontologies and creating relations among them using
bridge files
Researchers should contact the curators before using the IDOPlant, because it is under active development. A draft is available at http://purl.obolibrary.org/obo/idoplant.owl. 3.1
Our review of the IDO suggests that it is generally
appropriate as a foundation for the description of plant diseases. The
IDO is rooted in the Basic Formal Ontology (BFO)
The IDO consists primarily of terms specific to
infectious disease, together with relevant terms imported from
other ontologies, such as organism from OBI; disease,
disorder, and disease course from OGMS; habitat from
ENVO; macromolecular complex, reproduction, and entry
into host from GO; bacterium and virus from NCBItaxon;
and molecular entity from ChEBI. The IDO has created
many new terms, such as resistance to drug, infectious
agent, and infectious disease epidemic. The bulk of the
unique IDO terms can be used for the IDOPlant without
modification. For example IDO:infectious disease is defined
as “A disease whose physical basis is an infectious
disorder”. This in turn is based on the OGMS definition of
disease: “A disposition (i) to undergo pathological processes
that (ii) exists in an organism because of one or more
disorders in that organism”
IDO terms such as transition to clinical abnormality
or subclinical infection required careful consideration,
because the word “clinical” is not commonly used for plants.
We decided that the meaning of their definitions was
appropriate for plants, despite the names. For example, a feature
of an organism is clinically abnormal when it: “(1) is not
part of the life plan for an organism of the relevant type …
(2) is causally linked to an elevated risk either of pain or
other feelings of illness, or of death or dysfunction, and (3)
is such that the elevated risk exceeds a certain threshold
level”
Another potential limitation of the IDO for use in plant science is the meaning of terms from the OGMS that were defined within the scope of clinical encounters involving humans. In particular, the definition of symptom from OGMS requires a host of a type that can report its experiences, and so is restricted to sentient hosts. In plant pathology, “symptom” is commonly used to describe the phenotypes that are associated with a plant disease. The phenotypes are independent of the disease and the same phenotype or “symptom” may be associated with many different diseases. Furthermore, diagnosis generally depends on a collection of phenotypes, and not every instance of a disease will display every phenotype that is typical of the disease. Rather than use the OGMS definition of symptom, we developed a new term for the IDOPlant: plant disease symptom =def. A feature of a plant that is of the type that can be hypothesized to be involved in the realization of a plant disease.
Comment: Features include phenotypes such pale yellow leaf color, processes such as sudden wilting, and independent continuants such as leaf lesion.
The terms plant disease symptoms already exist in other ontologies (primarily the TO), and will be linked to plant diseases using the relation has_plant_disease_symptom (see section 3.3). 3.2
The study of plant diseases encompasses many
domains. In addition to IDO terms common to all infectious
diseases, like pathogen or resistance, the IDOPlant needs
terms to describe the taxonomy of host plants, pathogens,
and vectors, genomic and genetic data, the geographic
location and ecology of diseases and hosts, plant and fungal
anatomy, plant and pathogen development, biological
processes, and molecular functions. To encompass this range,
the IDOPlant is not only creating new ontology terms
specific to its domain, but also integrating and linking to
existing terms from multiple sources (Table 1). Whenever
possible, existing ontology terms are being used to create logical
definitions for IDOPlant terms. For example, rice bacterial
leaf blight is defined as “A bacterial blight disease (in
IDOPlant), that has as infectious agent Xanthomonas oryzae
(from NCBItaxon)” (fig. 3). Terms from external ontologies
will also be used for relations such as rice bacterial leaf
blight disease has_plant_disease_symptom pale yellow leaf
color (from TO). Logical definitions allow us to use
automated reasoners to ensure that the ontology hierarchy is
sound and to infer sub-types and relations implied by the
definitions. These can then be added to the ontology
if correct or eliminated if incorrect or redundant
In addition we are developing the following for IDOPlant: has_plant_disease_symptom: This relation is used to indicate a phenotype, process, or independent continuant that is evaluated to diagnose a disease. For example, “rice bacterial leaf blight disease has_plant_disease_ symptom leaf color pale yellow” means that pale yellow leaf color is a plant disease symptom (see above) of rice bacterial leaf blight disease, but it does not mean that every instance of rice bacterial leaf blight disease has pale yellow leaves. 3.4
Much of the information available on plant diseases is in a natural language or free text form, such as: “Bacterial leaf blight disease of rice is caused by Xanthomonas oryzae. It produces pale yellow leaves in mature plants. In one report the pathogen and the disease were reported in the Northern Territory of Australia.” Using ontologies to process such descriptions in a standardized form makes them comprehensible to computers and reasoners. For example, the description above could be converted (using natural language processors or other mechanisms) to: disease: rice bacterial leaf blight disease | host species: Oryza sativa (rice) | caused by: Xanthomonas oryzae | has symptom: pale yellow leaves | reported in: Northern Territory This standardized text could then be made even more powerful using ontology terms and relations (Fig. 3).
The current situation in the plant disease research
community is similar to that in the animal community when the GO,
MeSH
The IDO imports the Relation Ontology (RO)
consideration by the IDO, is used to indicate Fig. 3. Some of the terms and relations needed to model rice bacterial blight disease in the IDOPlant. Following the IDO, a disease is treated as a disposition of an infected organism, which has a particular infectious agent. The IDOPlant can also be used to define terms in the TO, such as rice bacterial blight disease resistance, which is a resistance to infectious disease that inheres in Oryza sativa. projects were being initiated: A number of organismspecific databases are faced with large amounts of molecular, germplasm (stock), genotype, and phenotype data associated with function, phenotype, or environment. The sharing of the task of building a set of controlled vocabularies such as GO and PO has helped enormously to address the needs of multiple individual databases. The IDOPlant controlled vocabulary for plant infectious diseases will allow database curators to store and retrieve the results of experiments related to diseases, including quantitative trait loci, pathogen and host germplasm descriptions, microarray expression studies, gene knockouts, reporter gene expression patterns, and gene-gene interactions from host and pathogen. The Plant Phenotype and Stress Ontology Project aims to overcome the obstacles in annotating data for complex biological concepts that span multiple ontologies by developing both the ontology terms and the software tools needed to annotate data from all aspects of plant diseases.
To annotate plant infectious disease description data, the IDOPlant is reaching out to resources such as the Food and Agriculture Administration’s AGROVOC (http://aims.fao.org/website/AGROVOC-Thesaurus/sub) and the International Rice Research Institute (http://www.knowledgebank.irri.org/rice.htm). These resources will enrich the IDOPlant by providing a wealth of information that can be incorporated into the ontology and by identifying gaps and errors. The IDOPlant can benefit these organizations by making their content more easily accessible to semantic applications. 4
As the growing human population and climate change place even more uncertainty on food supply, the need to understand the linkages between plant disease, environment, and yield is greater than ever. The IDOPlant and the Plant Phenotype and Stress Ontology Project can contribute to this challenge by making data on plant diseases more accessible. We are taking advantage of the interoperability of OBO Foundry ontologies to leverage existing terms to enhance the new IDOPlant extension, simultaneously enriching all ontologies involved by filling in terms needed for logical definitions. By expanding the core terms of the IDO to plants, we can learn how plant diseases differ from, and are similar to, infectious diseases in general.
RW, JE, DWS, and PJ are supported by NSF-IOS: 0822201 to the Plant Ontology Project. BS is supported by NIH:U54 HG004028 (National Center for Biomedical Ontology) and NIH / NIAID R01 AI 77706-01 (Immune System Biological Networks). Thanks to Lindsay Cowell, Laurel Cooper, Robert Hoehndorf, and three anonymous reviewers for comments on earlier versions of this manuscript.