nstitute an experimental study. We illustrate the challenges using real-world examples from a complex neuroscience integration problem and present the methodology and some of our tools, in particular the KIND mediator prototype for model-based mediation of scientific data. 6.1 Background Seamless data access and sharing, handling of large amounts of data, federation and integration of heterogeneous data, distributed query processing and application integration, data mining, and visualization are among the common and recurring broad themes of scientific data management in many disciplines. A main stream of activity in the bioinformatics domain is concerned with sequence and structural databases such as GenBank, NCBI, PDB, SwissProt, etc. and much work is devoted to algorithmic challenges stemming from problems, e.g., e#cient sequence alignment and structure prediction. However, in addition to the well-known challenges of main stream bioinformatics applications such as algorithmic c