An ontology for microbial phenotypes

OMP structure and organization

OMP is a general microbial phenotype ontology applicable to diverse microbes. Microbial phenotypes include properties of single organisms as well as populations of cells, such as colonies, biofilms, pellicles, and liquid cultures. Entities described by OMP include phenotypes that are directly observed and phenotypes that are inferred from indirect evidence. Examples of directly observed phenotypes are cell and colony morphologies and the presence or absence of motility. Inferred phenotypes include most physiological phenotypes. For example, while it is technically possible to observe utilization of a sugar by isotopic tracer experiments, it is much more common to infer that utilization is present or absent based on the ability or inability of the sugar to support growth or transform indicator dyes.

Furthermore, there are two distinct classes of phenotypes: independent phenotypes that provide information about an organism without reference to any other information and dependent phenotypes that capture a relative difference observed when comparing (at least) two separate genotypes or conditions. To illustrate the difference between independent and dependent phenotypes consider the phenotypes associated with motility. If a strain of E. coli were observed to move, it would have the independent phenotype “presence of cell motility”. However, if a mutation in a particular E. coli strain led to a reduction of motility, then that strain would be described by the dependent phenotype “decreased cell motility,” which is relative to the strain used as a reference. Organisms that are non-motile, such as Klebsiella pneumonia, would have the independent phenotype “absence of cell motility” [3]. Terms for both independent and dependent phenotypes are essential to represent the richness of genetics and genetic interactions (for examples, see [18]-[22]). OMP and its associated annotation system have been designed to capture all of these types of phenotypes. In the ontology, terms for independent and dependent phenotypes can be distinguished by their definitions as well as by the fact that dependent phenotypes are linked to relevant parent class “altered phenotype” terms. We have examined the term structures of related ontologies such as the Fission Yeast Phenotype Ontology (FYPO) [14] in order to ensure compatibility between OMP and other resources. We hope to ultimately develop OMP to be able to represent all phenotypes in other ontologies related to microbes allowing analysis of annotations to be linked together into one system.

OMP is built using foundations provided by Basic Formal Ontology (BFO) [23] and Phenotype and Trait Ontology (PATO) [24]. BFO provides a framework of upper level terms to support domain-specific ontology development. PATO contains phenotypic attributes for use in developing phenotype terms. Figure 2 shows how OMP relates to BFO and PATO, using terms related to motility as an example. All OMP terms (classes) descend from a root class called “microbial phenotype”. Because microbial phenotypes describe both processes (e.g. motility) and objects (e.g. cells), the OMP root class would be a direct child of the BFO root class “BFO entity.” However, the more granular (specific) OMP classes also connect to either the “continuant” or “occurrent” branches of BFO [25] by way of external ontologies, such as PATO or GO, as shown in Figure 2. Continuants are entities that persist through time, and include both objects, such as a bacterial cell or colony, and qualities of objects, such as cell shape or colony color. Occurrents are entities that have a temporal part and that occur through time, and include processes that an object can participate in, for example cell motility. In order to qualify processes (e.g. “decreased cell motility”), OMP follows the model of PATO, which defines “process quality” (PATO:0001236) as “a quality which inheres in a process.”

We have included umbrella terms to group independent and dependent phenotype terms that refer to the same biological concepts (Figure 3). For example, the terms “presence of cell motility”, “absence of cell motility”, and “decreased cell motility” are all related to the same biological concept (cell motility) and are all thus also subclasses of “cell motility phenotype”, itself a subclass of “locomotion phenotype” (Figure 2). “The umbrella term names end in the word ‘phenotype’ (Figure 3).” These terms are particularly useful for querying of related annotations. We recommend that they not be used directly in annotations as there will usually be a more specific child term available. These terms are similar to the Gene Ontology (GO) grouping terms such as “cell part”, GO:0044464, which GO guidelines indicate should not be used for annotation [26],[27].

Finally, OMP employs the qualifier “altered_relative_to” for dependent phenotype terms, which denote a difference in phenotype between organisms of two different genotypes or the same genotype assessed in two different environments. We have avoided the use of the qualifier “abnormal” for two reasons: there can be no consistent definition of what is “normal” or “abnormal” across organisms, and “altered_relative_to” is superior for capturing phenotypes of genetic modifiers such as suppressors and enhancers [28]. An altered phenotype is always relative to the phenotype of a designated control, which can be specified in the annotation (see below), or set to an arbitrary reference within an annotation set (e.g. a wild-type reference genotype within a species).

Term development

All terms in OMP have genus-differentia definitions, where a term is a subclass of a parent term and is distinguished from other related subclass terms (siblings) by some differentiating characteristic unique to that term (and could be phrased as “B is an A that C’s”). For example, “cell motility by gliding” (B) is a “presence of cell motility phenotype” (A) that has the differentia “where a cell or cells exhibits smooth movement along a solid surface” (C). Terms also contain synonyms that allow for more effective keyword searches to be conducted. For instance, the OMP term “serotype” has synonyms “antigenic typing phenotype,” “serological test phenotype,” and “serovar.”

In addition to standard definitions and synonyms, many OMP terms contain a cross product, a special type of logical definition that represents the intersection of two or more terms [29]. These are based on a specified genus and a collection of differentiae. OMP cross products are derived by using relationships to connect a term from within the ontology to terms from external ontologies. For example, the OMP term “motility phenotype” is equivalent to the intersection of the OMP root term “microbial phenotype” (OMP:0000000) and the GO biological process term “cell motility” (GO:0048870) using the OBO Relations Ontology [30] relationship “inheres_in” (RO:0000052). In OBO syntax, this logical definition is expressed as:

OMP:0000001

name: cell motility phenotype

is_a: OMP:0000312 ! locomotion phenotype

intersection_of: OMP:0000000 ! microbial phenotype

intersection_of: inheres_in GO:0048870 ! cell motility

intersection_of: inheres_in GO:0005623 ! cell

Cross products are also used in OMP to implement the entity-quality (E-Q) method of phenotypic character description. In the E-Q model a bearer entity (e.g. a cell, cell part, or biological process) is associated with a quality (e.g. shape, decreased number, or altered duration), and together these comprise the “phenotype” [31]. OMP uses entity terms from OMP or other appropriate ontologies such as the GO along with qualities from PATO. For example, the term “decreased cell motility” is a type of “altered cell motility,” with the added quality “decreased process quality.” In this example, “altered cell motility” comes from within OMP and “decreased process quality” comes from PATO. The two terms are joined by the relationship has_quality (RO:0000086). In OBO syntax, this logical definition takes the form:

OMP:0000002

name: decreased cell motility

is_a: OMP:0007001 ! altered cell motility

intersection_of: OMP:0007001 ! altered cell motility

intersection_of: has_quality PATO:0002302 ! decreased process quality

Whenever OMP term logical definitions comprise terms from external ontologies, OMP policy is to work with those developers to achieve interoperability.

Most of the existing OMP terms were generated as the result of OMP working group meetings held at Texas A&M University (College Station, TX), the Institute for Genome Sciences (Baltimore, MD), or by conference call. The term generation process includes examination of texts such as Bergey's Manual of Determinative Microbiology[3] and microbial literature, particularly genetic papers and reviews that are enriched for phenotype information (e.g. [1],[22],[32],[33]). The process of establishing the structure of OMP and building the initial term set was carried out over several months during the beginning stages of the project. Since then, we have periodically refined the structure, as needed based on feedback from the community and project annotators. Now that annotation using OMP has begun in earnest (see more information below) we are able to engage in annotation-driven term development in which annotators identify phenotypes of interest from publications and request new terms as needed. This allows us to ensure that the ontology contains the terms and structure needed to fully capture real-world information as it will be encountered by annotators outside of our group. Ongoing development of OMP is in response to OMP literature curators, focusing on the genetic literature of the well-studied model organism, E. coli K-12. Following the model of the Gene Ontology [26] and other National Center for Biomedical Ontology ontologies [34], OMP is designed to support community participation in term development through the use of a public tracker [35].

OMP is developed in Open Biological Ontologies (OBO) format syntax with the ontology editor OBO-Edit [36]. A term tracker [35] is used to manage term requests, which are researched and added into the ontology. The built-in OBO-Edit reasoner is run to check term relationships and validity of cross-product terms.

Accessing OMP

The first version of OMP was released in June 2011 on the OMP project’s development site [17], and new versions are released on a regular basis. Both a developer’s version and official release versions of OMP can be viewed at or downloaded from the development site. As of this writing there are 724 terms in the OMP name space, of which 100% have definitions. We have also deployed a wiki [16] modeled on Gene Ontology Normal Usage Tracking System (GONUTS) [37] for exploring the ontology, and adding usage notes to terms. The preliminary annotations described above are publicly visible via the OMP wiki. This will be converted to a system for community contribution of phenotype annotations in the future. Users wishing to request additions or changes to OMP may do so using the issue tracker system at our development site [35].

Use of OMP to capture phenotype annotations

OMP can be used with any user-defined annotation system to capture phenotypes in a set of appropriately defined fields containing accessions or free text. OMP could be used in single-species annotation systems similar to the ones currently used at Saccharomyces Genome Database [38] and PomBase [39], where Yeast Phenotype Ontology and Fission Yeast Phenotype Ontology terms, respectively, are associated with gene records. However, to fully leverage the design of OMP we are developing an annotation data structure that will provide a more general and powerful mechanism to store both independent and dependent phenotypes associated with specific genes and alleles. A key element of this system will be identifiers associated with each annotation that will allow dependent phenotypes to be more fully expressed using relationships between annotations. Figure 4 shows two types of annotations: Figure 4A depicts an independent phenotype annotation while Figure 4B shows how a dependent annotation is made relative to the independent annotation in Figure 4A.

While the more complex annotation system is under development, we are continuing to collect phenotype information from publications on E. coli and storing the following elements: an identifier for the strain or genotype being annotated, a description of the environment or experimental conditions, an OMP term to capture the phenotype, an Evidence Ontology (ECO) term [7] to capture the evidence that supports the annotation, and the reference that was curated to make the annotation. So far we have associated phenotypes to observations made in 275 papers. These annotations connect OMP term development to the corpus of papers used to find examples of phenotypes that OMP must be able to support. Annotations can be viewed on the OMP wiki.