The biological data for the proposed database system is already being generated by separately funded research projects in our labs and other zebrafish labs who have agreed to contribute data (letters, Section I). Considerable resources, however, are required to plan, design, and implement a database system that will aptly serve the entire research community (see Budget).
To judge the cost-effectiveness of this particular solution, we can consider at least these three alternatives: 1) Do nothing at this point. 2) Establish several lower cost data servers, like the existing Mosaic server. 3) Establish a more traditional, single centralized system either by adapting an exiting system, like ACeDB, or by creating a new database system and custom GUI.
Alternative (1) is completely unreasonable for the reasons outlined in the Problem Statement (above) including the amounts of data that are being generated at an increasing pace. Without adequate means for catagorization, storage, and retrieval, much of these data will be lost or forgotten. Neither solutions (1) or (2) will ensure data compatibility among different sites and neither provides true querying capabilities or a realistic way to make global searches among different data types at different locations. Option (3) precludes the advantages of distributing the databases to the sites generating data. Moreover, for world-wide access, we would need to write, distribute, and maintain specific client software for all the platforms used by the zebrafish research community to access the GUI. Although adapting an existing system might be cheaper than our proposed solution, none of the existing systems support true relational queries and none incorporate images with other data types. Most existing systems use X-terminals for the client software, whereas the majority of zebrafish labs use Macintosh or PCs and do not have access to UNIX machines. Finally, to ensure consistency in the data types, considerable funds will be needed to develop a common data model anyway; so why not implement that data model in a state-of-the-art database system?
Continue on to the Bibliography