BIRN-M: a semantic mediator for solving real-world neuroscience problems

A goal of the Biomedical Informatics Research Network (birn) project is to develop a multi-institution information management system for Neurosciences to gain a deeper understanding of several neurological disorders. Each institution specializes in a different subdiscipline and produces a database of its experimental or computationally derived data; a mediator module performs semantic integration over the databases to enable neuroscientists to perform analyses that could not be done from any single institution’s data. The overall system architecture of the birn system is that of a wrapper-mediator system. The information sources are various relational sources including Oracle 9i having userdefined packages, Oracle 8i with the Spatial Data Cartridge, and databases made available over the web. Sources also include computational resources that need to be “run” to produce data. A source can be either accessed directly by the mediator, or through a middleware called the Storage Resource Broker (SRB) to shield the mediator from keeping low-level information like data location and user authentication information, and from large object handling. Wrappers handle source registration, schema mapping, and translation from the mediator’s logic queries to the corresponding sql or web request. birn-m also admits knowledge sources, which host ontologies and domain-specific general knowledge in the form of logic rules. birn-m maintains a registry with semantic schemas from all sources, integrated views defined in a GAV fashion, and source query capabilities as binding patterns and special predicates or functions admissible by it. The registry also records the types of data a client can handle, and has an API for interactive custom-built clients.