Heterogeneous Data Integration with Mobile Information Managers

Genomic and protein sequences, and in general molecular biology data, belongs to a relatively new information medium, with its own specific processing algorithms. Integrating such data from heterogeneous sources presents many of the challenges presented by integrating more common multimedia database integration. In this talk we are discussing solutions built as distributed applications because of the need for information sharing and that are easily scalable and reconfigurable in order to accommodate the frequent changes requested by molecular biology researchers. Specifically, we consider Mobile Information Managers (MIMs) which combine a specific approach to generating mediators with some aspects of agent technology, with everything based on a strongly uniform formal foundation. MIMs are downloadable, networked components that have to a large degree uniform capabilities and can run at any site that implements the execution engine (virtual machine) for them. The basic functionality of a MIM is to implement a data integration and transformation from one or more data sources to one (virtual) data target. This target is therefore an (unmaterialized) view of the sources, and it is offered as yet another data source to the rest of the application. Thus, MIMs communicate with queries flowing in one direction and data returning in the other A MIM will contain a high-level description of the schemas for sources and target and of the integration/transformation that governs its activity. These descriptions, as well as the description of the queries and the data are formalized in the same internal framework.