What is BIRNLex?

BIRNLex was created as a controlled terminology for annotation of BIRN data sources. These sources currently include image databases, containing data from structural and functional magnetic resonance imaging (MRI) on human subjects involved in studies on Alzheimer?s disease and schizophrenia, and multiscale imaging of mouse models of human neurological disease using MRI, light and electron microscopic imaging. The BIRNLex provides terms, utilized by BIRN scientists in the context of their research, covering neuroanatomy, molecular species, behavioral and cognitive processes, subject information, experimental practice and design, and associated elements of primary data provenance required for large-scale data integration across disparate experimental studies. Many of the terms are drawn from existing terminologies and ontologies such as the Unified Medical Language System (UMLS), the Foundational Model of Anatomy (FMA), Neuronames, the Gene Ontology, the Ontology of Biomedical Investigation (OBI) and the Phenotype and Trait Ontology (PATO), and, along with these constituent ontologies, BIRNLex will be made publicly available to search and browse via the NCBO BioPortal. These terms will form the basis of a more formal ontology describing the multiscale investigation of neurological disease - the MIND ontology.

The BIRNLex is maintained by the BIRN Ontology Task Force (OTF). All members of the BIRN community are welcome and encouraged to contribute to its construction and curation.

Structure of BIRNLex:

Each term in the BIRNLex is accompanied by a human readable definition, a unique identifier, the curation status and curation date, the contributor of the term and the source of the identifier. Synonyms are also identified when available.

• Definitions: The definition of the term as it is used in BIRN is given in the BIRN definition field. As much as possible, the definition of each term adheres to the form "A is a B which", e.g., "an electron microscope is a microscope utilizing electrons and magnetic lenses to form an image." Defining terms this way as qualified cases of higher order classes makes it easier to place each term in its proper place in the hierarchy and derive the complete set of attributes for each term when constructing formal ontologies from the BIRNLex. If the definition was obtained from another source, the source should be acknowledged.
• Unique identifier: Each term is accompanied by a unique identifier termed the Bonfire ID. The Bonfire repository contains knowledge sources collected by BIRN and utilized by the BIRN data integration environment to resolve concept-based queries. The Bonfire repository contains the current version of the UMLS and Neuronames and can be extended with additional vocabularies as necessary. The BIRNLex contains a subset of the terms contained in these vocabularies. These are selected to cover all terms in the BIRN data collections and other terms that are closely related to them. The unique identifiers reflect the source of the term, e.g., the UMLS CUI (Concept Unique Identifer), so the term can be cross referenced with its source terminology. The prefix "BF" in a Bonfire ID indicates the term was added by a BIRN member to Bonfire and may not necessarily have been derived from a pre-existing terminology/ontology.

The following metadata attributes are associated with each BIRNLex term and used to track the provenance and evolution of the terminology:

• Curation Status: This required field must contain one of the following values: ?U? for uncurated, ?C? for curated and vetted, and "C-star" for curated but requiring additional review. Curation is performed by the BIRN Ontology Task Force (OTF) with participation by the BIRN community. The OTF organizes curation sessions several times per year in which each term is considered and modified as necessary to ensure it contains a clear, crisp definition conforming to the BIRNLex standards. Curation is an ongoing and iterative process. The designation ?U? connotes the term has not yet been considered for curation but in no way indicates the term is viewed unfavorably or of lesser status than any other term in the lexicon.
• Curation date: This field indicates the date the term was curated.
• Contributor: This field specifies the individual or organization responsible for contributing the term to the lexicon. The contributor may or may not be the originator of the term.
• Source: The source terminology/ontology from which the term and its unique identifier were derived, e.g., UMLS. If this field is left blank, then the source is usually the contributor. Over time, we will associate a unique identifier with every BIRNLex term acquired from an external source. The format of this identifier will conform to that of a Universal Resource Identifier (URI) recommended by the WWW consortium (W3C) as web object pointers. The first benefit of this will be to help the user navigate the URI link to find all other information provided by the external source about the term. Down the road, it would also help us treat every URI-identified term as an RDF resource, and thus will make our representation compliant with the semantic web standards of the W3C. Consequently, we will be in a position to use tools developed by the semantic web community to search through and reason with large terminologies.

Building the BIRNLex

The BIRNLex began as an outgrowth of an Ontology Workshop held in San Diego in January 2006. At this workshop, participants were asked to identify the subset of Bonfire terms immediately relevant to BIRN data. If there was no existing term, participants added the term to the Bonfire repository and linked it to one or more existing terms. Each term added was given a Bonfire identifier. At the time, the Bonfire consisted largely of the UMLS. The UMLS was found to be an excellent source for establishing semantic equivalence among synonymous terms, e.g., Purkinje cell = Purkinje neuron. However, difficulties arose because many terms in UMLS had no definitions, the relationships and semantic types used to group terms into higher-level categories were inconsistent. The UMLS is also very large, containing hundreds of thousands of terms, the overwhelming majority of which were not required to build KR Maps for BIRN specific data. Furthermore, it was not possible to assign new attributes directly to terms as they exist in UMLS. Based on advice from Dr. Daniel Rubin of the National Center for Biomedical Ontologies (NCBO), the BIRN OTF decided to create a simplified lexicon, modeled after the Radiology Lexicon RadLex, to define the terms necessary for BIRN participants to annotate their data. These terms would then be used to construct a more formal ontology covering the multiscale investigation of neurological disease.

The OTF began construction of BIRNLex using the terms gathered during the January 2006 workshop. The OTF then attended an Ontology Workshop at Stanford University at NCBO, hosted by Drs. Mark Musen, Daniel Rubin, Suzanna Lewis and Barry Smith. As experts in biomedical ontology construction and use of controlled terminologies for large-scale biomedical data annotation, these researchers were asked to review the current state of BIRNLex and the BONFIRE repository. Based on feedback from NCBO, the OTF established a set of best practices for constructing the BIRNLex. As a starting point for describing and organizing terms related to experimental practices employed across all BIRN testbeds, we adopted the the Ontology of Biological Reality (OBR - http://ontology.buffalo.edu/medo/biomedo.htm) used in FUGO (www.fugo.sourceforge.net), the FMA, and being applied to other biomedical ontologies as a foundational structure, as well as the OBO relations ontology (http://obo.sourceforge.net/relationship/) as a source of term relationships. We also selected specific, assay-related terms from FuGO, adding additional entities specifically related to neuroimaging. Because the BIRNLex is not limited to experimental practices, terms were also added for gross and cellular neuroanatomy, cognitive and behavioral processes, molecular entities and taxonomic species and strain. For the cognitive and behavioral terms, the OTF collaborated with the BrainMap group in San Antonio. The current BIRNLex is not meant to be comprehensive, i.e., it does not contain comprehensive lists of terms from any given domain. Rather, the goal of BIRNLex is to provide sufficient coverage to describe the current BIRN data. Furthermore, BIRNLex will be evolving and growing in the future to provide more granular and more comprehensive descriptions of the data in BIRN and in the neuroscience domain.

As each term is added, we attempt to place it in the lexical hierarchy, based on the definitions supplied. For those biomedical domains possessing mature ontologies, relatively little effort was made to place terms into detailed hierarchies, as BIRN expects to derive these relationships from the associated external knowledge resources. For example, because several knowledge resources exist for neuroanatomy, we did not attempt to recapitulate the detailed classification existing within these resources, because these are already available through the Bonfire repository. For domains not currently covered by mature knowledge resources, e.g., experimental practice and cognitive and behavioral processes, the BIRN OTF and other BIRN personnel are devoting more effort toward filling out the domains. However, because BIRNLex is intended to be a lexicon rather than a comprehensive ontology, terms are added even if the hierarchy isn?t clear. We anticipate creating the necessary hierarchies for experimental practice and complex phenotypes will involve extensive interaction with the FUGO and PATO groups and others developing ontologies intended for use in creating formal, detailed specifications of biomedical experimental data provenance.

What does BIRNLex cover?

The BIRNLex is meant to cover those domains of biomedical reality required for the annotation of BIRN data. Because the BIRN is generating a large number of images, and images require a context for proper interpretation, the BIRNLex will contain a large number of terms describing experimental techniques. Other domains include (but are not limited to):

• Anatomy
• Imaging , e.g., imaging devices, imaging modes and parameters
• Experimental design
• Protocols
• Project administration, e.g., project, experiment
• Organism taxonomy ? e.g., species and strain (for mouse)
• Data types and data provenance
• Phenotypes:
  • Behavioral, cognitive, anatomical, biochemical, molecular
• Environment

Contributing to BIRNLex

All members of the BIRN community and their collaborators are welcome and encouraged to contribute to BIRNLex. Terms may be contributed either by e-mailing the Ontology Task Force or through the BIRN Ontology Wiki . Each term should be accompanied by a definition in the form ?A is a B which??. Here are two example definitions:

• basal forebrain: region of the brain consisting of ventral and rostral subcortical sub-regions of the telencephalon, including the basal ganglia, septal nuclei, amygdala, ventral pallidum, substantia innominata, and basal nucleus of Meynert.
• Electron microscope: A microscope utilizing electrons and magnetic lenses to form an image.