The following gives a brief description of the components of
BrainML. A more detailed
description is also available, as is an overview of BrainML's
purpose and architecture.
BrainMetaL provides a basic set of tags that are
applicable to most areas of neuroscience (as well as many areas
outside of neuroscience). Most tags can only be used in one way
under the rules of XML Schema. Some others though can be used
in multiple ways, but should be used in particular ways
in order to support processing by BrainML-aware software. In
some cases the software will enforce these conventions, while
in others a degree of latitude is allowed in order to avoid
limiting the data exchanged using BrainML or the processing
model that applications use.
Lists of controlled vocabulary terms may be created in XML
following a particular schema. The terms are organized by
ISA hierarchy and domain of applicability. Models may
specify that particular fields are to take their values from
controlled vocabulary from a particular domain of
As with controlled vocabulary, lists of units may be created
in XML following a particular schema. There is no ISA or
domain of applicability organization in this case. (In a
future version applicability specification by dimension may
be added.) Models may specify that particular fields are
to take units.
In addition to inheritance and aggregation, models may
declare that entities are related to one another without
specifying anything further about the relationship. In
this case, a special <link> element
is placed in instance documents as a child of the "from"
end of the link, naming the "to" end by XML ID.
BrainMetaL defines a default bibliographic citation format
suitable for journal articles, book chapters and proceedings
papers, books, and theses. If it is desired for some
reason to use a different format, an external schema may be
referenced and its format used inside a special "wrapper"
BrainMetaL defines general purpose containers for one- and
multi-dimensional (rectangular and non-rectangular) data.
The available representation choices are decimal, integer,
and string. The available formats are tagged (full
structure in the XML), compact (comma-separated values and
similar), and binary (Base64-encoded blocks). Decimal
numbers may be arbitrary precision, however applications are
not guaranteed to support any more than IEEE double
precision. This is also the format the binary data uses for
decimal. Likewise, for integers, application and binary
support is for 64-bit integers.
While it is neither possible nor desirable to provide an
exhaustive data model for all of neuroscience, it
is possible to provide a basic set of very general
categories under which other entities can be defined. This
basic ontology is helpful in providing some minimal
structure that can serve as an organizational seed for both
BrainML models and applications. BrainMetaL provides this
seed in the form of the Quintessence definitions,
five top-level basic categories relating generally to
neuroscience data: data, entity,
reference, method, and model.
BrainML (Base Distribution)
BrainML is the extensible system of XML schemas built
on top of BrainMetaL (i.e., using BrainMetaL tags) together with
conforming instance documents. A set of additional tag
definitions is provided with the BrainML specification to aid in
building neuroscience models. Unlike some of the BrainMetaL
parts, the use of these is governed strictly by XML Schema
rules rather than convention.
Experiment / View / Trace
BrainML defines structure for data representation over and
above that defined in BrainMetaL. A set of tags is provided
for representing an experiment as a set of one or
more views each composed of one or more
traces (which themselves may be single or
multi-dimensional). This tag also acts as a generic package
for a data submission (to a repository) containing the tags
Recording Site / Source / Location
BrainML defines tags for representing the recording site of
a neurophysiology experiment. These are defined at a
general level, and are expected to be refined by specific
data models for actual use (see, e.g, the cortical
model. The motivation for the general tags is to allow
applications a means of recognizing recording sites
generically even when the details actually used for
representation can change.
Similarly to recording site, a tag for experimental Protocol