Editorial: Building successful biological databases

In this issue, we are happy to present a collection of four articles that discuss the key factors in building a successful biological database. The bioinformatics community does many things, but we can roughly summarise most activities as either building algorithms or building databases. Successful algorithms are measured in terms of their time and space complexity, and there are well-known standards for communicating and validating new algorithms. Databases are much more difficult to validate. How do we know that a database is good? Do we judge it by the quality or quantity of its data, its usability and its ability to integrate with outside resources? Should we also consider the technical architecture and the user support mechanisms and documentation? What about the impact on science, the number of web hits, and total gigabytes of data that a database transfers? Of course, we probably should include all these factors in deciding if a database is a success. Anyone who builds and maintains biological databases also knows that it is very hard to publish traditional academic papers about databases. Familiar refrains include 'What's the hypothesis?', 'What's the evaluation scheme?', and 'This is not science, this is engineering'. In partial response to this difficulty, there is the annual special 'database issue' of Nucleic Acids Research, and the occasional Application Note in the back of the Bioinformatics journal, and other similar publications. But these articles are always severely limited in page length, and can sometimes read as database advertisements, written to attract potential users. We believe that Briefings in Bioinformatics has an opportunity to provide a forum for the maintainers of successful database projects to publish analyses of their approach, and to reveal the general lessons they have learned about the task of creating a successful biological database. Although many issues in creating a good database may transcend biology and be valid for all domains, there are special circumstances around biological databases that make them worth treating as a special group (eg the presumptions that they be freely available, that they be on the internet, that they keep up with a rapidly growing field and that they maintain high biological relevance). To assemble this special issue, we solicited manuscripts from the scientists associated with a shortlist of databases that the editors and their consultants felt could be called 'successful' with little controversy. These ranged over a variety of data types, but concentrated on databases …