Common data model for neuroscience data and data model exchange

Abstract

Objective: Generalizing the data models underlying two prototype neurophysiology databases, the authors describe and propose the Common Data Model (CDM) as a framework for federating a broad spectrum of disparate neuroscience information resources.

Design: Each component of the CDM derives from one of five superclasses-data, site, method, model, and reference-or from relations defined between them. A hierarchic attribute-value scheme for metadata enables interoperability with variable tree depth to serve specific intra- or broad inter-domain queries. To mediate data exchange between disparate systems, the authors propose a set of XML-derived schema for describing not only data sets but data models. These include biophysical description markup language (BDML), which mediates interoperability between data resources by providing a meta-description for the CDM.

Results: The set of superclasses potentially spans data needs of contemporary neuroscience. Data elements abstracted from neurophysiology time series and histogram data represent data sets that differ in dimension and concordance. Site elements transcend neurons to describe subcellular compartments, circuits, regions, or slices; non-neuroanatomic sites include sequences to patients. Methods and models are highly domain-dependent.

Conclusions: True federation of data resources requires explicit public description, in a metalanguage, of the contents, query methods, data formats, and data models of each data resource. Any data model that can be derived from the defined superclasses is potentially conformant and interoperability can be enabled by recognition of BDML-described compatibilities. Such metadescriptions can buffer technologic changes.

Figure 1

Top-level superclasses span neurophysiology. Each first-class component of the Common Data Model derives from one of five superclasses—site, data, reference, method, and model elements. Relations (shown as diamonds) provide links between elements and subclasses, including neurons, data sets, protocols, and publications.

Figure 2

Site element description for the somatosensory database. Site element abstracts neuronal anatomy, encompassing neurons, regions, and other anatomic structures. For cortical microelectrode neurophysiology, recording sites are primarily cortical neurons, defined here as subtypes of mammalian neurons. Attribute names ending in _CV have controlled vocabulary values. In this entity-relationship diagram, subclasses of site element are shown with names and data types of characteristic data and descriptive metadata attributes. Diamonds indicate relations between sites and other elements, including controlled vocabulary for neurotransmitter and receptor classes (not shown).

Figure 3

Molluscan neurons are described by a partially different scheme based on disparate techniques and community-based views of neuronal identity. Some descriptive attributes are the same as those used for cortical neurons and shown in Figure 2▶. Other attributes have distinct sets of values, and some attributes are unique to this class of neurons.

Figure 4

Samples of controlled-vocabulary values for each of four attributes. One (recording technique) is shared by cortical and invertebrate databases, one describes molluscan neuron receptive fields, and two are used as neuron descriptors and search terms in the cortical database. Hierarchies expand to the right, with broader terms shown to the left in each column. The hierarchic attribute-value search algorithm shown in Figure 6▶ enables selectable specificity.

Figure 5

Hierarchic attribute-value schema implements controlled vocabulary (CV) hierarchies compactly and efficiently. All CV values, the attributes they specify, and glossary definitions reside in a single terms table. The hierarchies table maps parent–child relationships, implementing vocabulary trees to any depth. As areas of investigation evolve and require enhanced specificity, hierarchies can be expanded without making prior entries obsolete. A sample attribute and value describing a cortical neuron receptive field are shown at the bottom right.

Figure 6

Hierarchic attribute-value (HAV) outer-match search algorithm enables selection of either operational or high specificity. The algorithm finds entries described by the search term or any of its more specific child terms. This approach favors the broad returns consistent with operational specificity yet allows high specificity searches when desired. The HAV schema is also compatible with exact-match search algorithms.

Figure 7

Biophysical description markup language (BDML) definition fragment for cortical neuron description. BDML serves as the human- and machine-readable Common Data Model interface. BDML is designed to describe the data formats, data model, and query terms for multiple data resources, mediating exchange by specifying commonalities. Based on resource description framework (RDF) and XML, BDML is independent of implementation and technology.

Figure 8

In a test of Common Data Model applicability to sequences of neuroimaging data, specification of time sequence enables trace animation. The canonical experiment > view > trace schema arranges multiple traces in a single view, ordered by the trace_seq index. To allow for animations, in which sequences of one or more traces are contained in a view, an extension to the schema is required. Deriving animated trace types from image traces by adding the additional attribute time_seq implements animation; values indicate temporal ordering of images. Parallel extensions can allow for animation of one- and three-dimensional traces as well. In similarly derived animated view types, the T_rate attribute specifies timing for constant-rate animation.

Figure 9

Schematic representation of multiple dimensions of interoperability for data-driven databases. User interoperability is discussed in a separate report on our two neurophysiology databases. Technical interoperability measures openness of architecture and utility of standards for data format specification and for data and data model exchange. Domain interoperability includes the scope of a resource and the ease with which it interfaces with resources representing different subfields or domains of a discipline. The data dimension measures relatedness of data and intersection of data models; the domain and data dimensions are thus non-orthogonal. Temporal interoperability reflects ease of migration and of incorporation of both future and legacy data.