The Subcellular Anatomy Ontology (SAO)
Download SAO 1.2
Browse the SAO 1.2 (on the NCBO Bioportal)
Download SAO 1.1
Browse the SAO 1.1
Download SAO 1.0
Browse the SAO
Special AnnouncementWe are greatly saddened by the sudden and untimely death of our friend and colleague, William Bug ("Bill"). Bill was a tireless contributor to the Biomedical Ontology community in general, and the Subcellular Anatomy Ontology in particular. He will be missed by all of us. The OBI community has set up a wiki where people can share tributes to our remarkable colleague. |
NEW! To view the CCDB / SAO Wiki, click here
NEW! To read a published article that describes the SAO from a neuroscience perspective, click here
Larson SD, Fong LL, Gupta A, Condit C, Bug WJ, Martone ME (2007) "A formal ontology of subcellular neuroanatomy", Front. Neuroinform. 1:3. doi:10.3389/neuro.11/003.2007
To view the SAO 1.2.5 on the NCBO Bioportal, click here.
To read a published article that describes SAO and its utilization, click here.
Fong LL, Larson SD, Gupta A, Condit C, Bug WJ, Chen L, West R, Lamont S, Terada M, Martone ME (2007) "An Ontology-Driven Knowledge Environment for Subcellular Neuroanatomy", OWL: Experiences and Directions, Innsbruck, Austria, CEUR Workshop Proceedings, ISSN 1613-0073, http://CEUR-WS.org/Vol-258/, June 6-7, 2007
To read a published article that uses SAO to perform rule-based reasoning, click here.
Larson SD, Martone ME (2007) "Rule-Based Reasoning with a Multi-Scale Neuroanatomical Ontology", OWL: Experiences and Directions, Innsbruck, Austria, CEUR Workshop Proceedings, ISSN 1613-0073, http://CEUR-WS.org/Vol-258/, June 6-7, 2007
Additional materials for the paper on rule-based reasoning are available here.
Background
We have developed the Subcellular Anatomy Ontology (SAO) for the nervous
system to provide a formal ontology to describe structures from the dimensional range known as the “mesoscale,” encompassing
cellular and subcellular structure, supracellular domains, and macromolecules.
The development of ontologies for neuroscience data is a key objective of neuroinformatic. An ontology consists of a set of concepts, or entities, within a domain linked by relationships such as “is a”and “has part,” e.g., “ neuron is a cell” and “cell has part plasma membrane.” Ontologies are highly valuable in that they provide a formalization of knowledge within a domain in a machine-readable form. Ontologies include a much wider scope of information than taxonomies, which are simply hierarchical representations of the concepts but lack formal descriptions of their properties and the types of relationships they have with one another. Ontologies have been used in many business and scientific environments to share, reuse, and process domain knowledge. A well constructed ontology supports the use of machine-based reasoning to derive new knowledge based on encoded relationships among individual data nodes.
Ontologies form one of the cornerstones of large
scale data sharing projects such as the Cell
Centered Database and
Biomedical Informatics
Research Network (BIRN) that are building
infrastructure for amassing and integrating biological data across dimensional
scales. Ontologies promote data
integration both by providing a common terminology for data annotation
and the means by which relationships among diverse data can be inferred
(Gupta et al., 2001 link
to PDF of paper by Gupta; should be on BIRN site).
As
part of the Cell Centered Database project, we are developing informatics
infrastructure for the dimensional range known as the “mesoscale”. This range roughly encompasses
the structures that sit between gross morphology and molecular structure,
e.g., cellular networks, cellular and subcellular microdomains
along with their macromolecular constituents.
These structures lie at the heart of information processing in the nervous
system, providing the adaptive spatial framework in which processes giving
rise to complex behaviors occur. The study of mesoscale structures continues
to present a challenge to experimentalists, because their dimensions fall
squarely between the capabilities of current imaging technologies. Investigations
of physiology, structural dynamics, coarse molecular distributions and
large-scale distributions of dendritic spines are typically accomplished
by optical microscopies. Appreciation of the fine structural detail
on internal structure, cytoskeletal organization, localization of molecular
constituents, location of synaptic contacts, and detailed views of the
immediate microdomain such as pre-synaptic boutons and glial processes
require 3D electron microscopic imaging. To build a comprehensive
understanding of the nervous system in this dimensional range requires
the ability to aggregate data obtained by multiple researchers across techniques
and spatial scales.
The Ontology for Subcellular Anatomy of the Nervous System (SAO) describes the parts of neurons and glia and how these parts come together to define supracellular structures such as synapses and neuropil (Fong et al., submitted). Molecular specializations of each compartment and cell type are identified. The SAO was designed with the goal of providing a means to annotate cellular and subcellular data obtained from light and electron microscopy, including assigning macromolecules to their approporiate subcellular domains. The SAO thus provides a bridge between ontologies that describe molecular species and those concerned with more gross anatomical scales. Because it is intended to integrate into ontological efforts at these other scales, particular care was taken to construct the ontology in a way that supports such integration.
Structure of the SAO
Cells are divided into regional parts, e.g., dendrite and axon, and component parts, e.g., mitochondrion, similar to the way the Foundational Model of Anatomy divides anatomical structures. Each of the parts of a cell can be further divided into regional parts and component parts, e.g., an axon can be divided into an initial segment and main axon. Each part of the cell is connected to its parent part through the relationships “continuous with”. Thus a dendrite is continuous with the cell somata; the dendritic spine is continuous with the dendritic shaft.Because the SAO is built on a model of the cell, both molecular constituents and anatomical location are assigned to the subparts of cells, rather than to the cell itself. The SAO utilizes the “located in” relationship to situate cellular parts into higher order brain regions. The SAO was constructed using OWL (Web Ontology Language), a first order description logic that supports reasoning. We have constructed the SAO in a way that allows inferencing to be performed across scales so that molecules and higher order connectivity may be inferred from local interactions (Larson and Martone, submitted ß link to PDF of rule based reasoning paper). Ideally, the observation that F-actin is located in the head of a dendritic spine from a Purkinje cell dendrite found in the molecular layer of the cerebellar cortex should allow for the following statements to be inferred.
The SAO directly builds upon several foundational ontologies recommended by the OBO Foundry Project, designed to promote the adoption of best practices in ontology construction to foster interoperability of ontologies with the broader biomedical community. At the most abstract level, SAO class structure follows the Basic Formal Ontology.
Class descriptions
Cell: The SAO contains a list of cell types found in the nervous system. Top level cell classes such as neuron and glia were taken from the Cell Type Ontology The SAO does not contain a comprehensive list of neuron types, because these entities fall under the scope of other ontologies. Rather, because the SAO is designed as an application ontology for annotation of biological data, the parent cell types are expected to be added to the SAO as they are encountered. The SAO lists neurons according to common names that reflect a mixture of classification criteria, e.g., morphology (``pyramidal neuron''), proper names (``Purkinje neuron''). The SAO utilizes these names merely as labels and does not further classify cell types into subtrees, except in instances where the hierarchy is fairly straightforward, e.g., layer 3 cortical pyramidal neuron is a cortical pyramidal neuron. We deliberately kept the cell classification flat because the SAO can be used to classify neurons along multiple dimensions through their specific properties, e.g., primary neurotransmitter, number of processes, anatomical location of cell parts.Part of Cell: The SAO comprises two main classes of cell parts: regional part and component part. Regional part of cell is elaborated under the BFO concept Fiat Object Part defined as a "part of an object not demarcated by any physical discontinuities ." Regional parts of neurons include dendrites, axons, the cell soma and protrusions such as dendritic spines. Regional parts of glia include the cell soma and processes such as astrocytic endfeet and myelinating processes. Each of these regional parts may be further subdivided into finer parcellations. Component parts are considered to be BFO independent objects and represent the building blocks common to all cells, e.g., organelles. Components are largely adapted from the Gene Ontology cell component hierarchy and cross referenced to the GO ID where possible, with additional neuron-specific components added where necessary. Macromolecules may be considered to be component parts, but are listed under a separate class in the SAO. Just as with cell types, the SAO does not contain an exhaustive list of macromolecules, because these entities are covered in other resources. As with cell types, we intend to keep SAO application-driven, and as such as molecules are encountered in biological data, they may be added to the SAO.
Supracellular structures: The SAO also includes a class termed ´supracellular structure¡. While somewhat paradoxical in light of use of ´subcellular¡ in the name of the ontology, this class refers to multicellular domains defined by subcellular parts of neurons and glia such as neuropil, synapses, and the Node of Ranvier. The supracellular designation refers to the fact that the subcellular parts are derived from at least two different cells. To classify supracellular domains according to the BFO, we used both the aggregate object and site classes. We consider some supracellular domains as aggregate objects because they represent a somewhat ad hoc grouping of cell parts into a higher order structures. which we believe best fits the BFO:ObjectAggregate defintion: "a mereological sum of separate objects possessing non-connected boundaries.". For example, the neuropil is a term applied to regions of the nervous system characterized by a dense tangle of intertwined cell processes each of which have distinct non-connected boundaries. Other supracellular structures are better characterized as sites, because they are believed to be the locations at which a particular function occurs. For example, the synapse is the site at which neurotransmission occurs. The location of that function is inferred because of the presence of one of more molecules or cell components involved in these processes. For some aggregate structures, we create an aggregate object and then a site where the object is located, e.g., the chemical synapse may be considered an abstract aggregate entity that consists of a pre-synaptic part, a post-synaptic part and a junctional part. Each of these parts have cell components, e.g., synaptic vesicles, located within them that define the extents of these parts. The site of the synapse is that part of the cell or cell-cell apposition where the parts of the synapse are localized. In this way, we restrict the synaptic site to that portion of the cell part occupied by synaptic components which certainly fits the BFO:Site definition as an entity possessing "a characteristic spatial shape in relation to some arrangement of other (entities) and ... (which) can be occupied by other (entities)".
SAO Properties: Properties in the SAO are grouped into part of, morphological and spatial relationships. Regional parts are assigned to each cell class using restrictions, e.g., neurons may only have neuronal regional parts. Each regional part is assumed to belong to a parent cell; geometrical relationships among cell parts are specified by relationships such as continuous with, e.g., dendrites are continuous with the cell somata. Although some properties are assigned at the level of cell class, e.g., morphological type, most are assigned at the level of regional cell part. In this way, cell components and macromolecules are assigned to the particular part of the nerve cell in which they are found. Similarly, because nerve cells are large and may span many brain regions, the property has anatomical location, designed to situate the cell within a regional part of the nervous system, is assigned separately to each part of the cell. Regional parts of brain will be drawn from other resources, e.g., BAMS
