Integration of Heteregeneous Bio-Medical Databases: A Federated Approach using Semantic Schemas

Biomedical experiments generate a vast amount of data that needs to be organized, integrated and analyzed. Important research challenges in the retrieval of these data include flexible, integrated analysis of data held in existing heterogeneous data sources. Indeed, the problems of supporting ad hoc queries across multiple data sources and correlating the data retrieved pose a host of challenging research problems. Our approach to integrate data from heterogeneous databases is to build a federated database system with a centralized mediator. We build a shared global schema and the mappings between the schemas are captured using rules. This paper focuses on issues concerning manipulation of large volumes of biomedical data in centralized, distributed or heterogeneous environments. We develop new computer science approaches to managing biomedical data, building on major biomedical informatics initiatives at Yale including over a decade of research performed as part of the national Human Brain Project. Both the functionality and performance of our system are being tested with data from the SenseLab database, CoCoDat database, and Cell Centered Database. ∗This work is partially supported by NSF grant 0331548. This work is also supported by NIH grant P01 DC04732. †Department of Computer Science, Yale University, CT ‡Department of NeuroBiology, Yale University, CT §Yale Center for Medical Informatics, and Department of Anesthesiology, Yale University, CT ¶Department of Computer Science, Yale University, CT ‖Center for Medical Informatics, Yale University, CT