The challenges and rewards of integrating diverse neuroscience information

The design of database models and schemas for storing, cross-referencing, and retrieving neuroscience information faces issues that are similar but more complex than most of the other biomedical disciplines, such as genomics and proteonomics. Specifically, the visualization and manipulation of very large and diverse image data, such as digital brain atlases and functional magnetic resonance images, play a unique role in neuroscience while much of the associated information is textually recorded. Nongraphical information can include the annotation of large brain structures ranging from anatomical regions to intracellular structures, the description of cellular functional properties, and their various interrelationships, such as fiber connections. It is necessary that the heterogeneous and distributed types of data be cross-referenced to each other so that this diverse information can be efficiently retrieved, shared, and exchanged among the different neuroscientific disciplines. Continued advances in computers and Internet technologies appear to indicate that increasingly large data sets will be maintained on local or regional file servers and that informational interoperability will be achieved using a networked information system infrastructure. The authors and others have proposed and implemented models of semantically organized information systems that utilize centrally stored and highly structured archival information to index, cross-reference, and retrieve diverse, Web-based data sets.