Database architectures for neuroscience applications.

Abstract

To determine effective database architecture for a specific neuroscience application, one must consider the distinguishing features of research databases and the requirements that the particular application must meet. Research databases manage diverse types of data, and their schemas evolve fairly steadily as domain knowledge advances. Database search and controlled-vocabulary access across the breadth of the data must be supported. We provide examples of design principles employed by our group as well as others that have proven successful and also introduce the appropriate use of entity-attribute-value (EAV) modeling. Most important, a robust architecture requires a significant metadata component, which serves to describe the individual types of data in terms of function and purpose. Recording validation constraints on individual items, as well as information on how they are to be presented, facilitates automatic or semi-automatic generation of robust user interfaces.