The Second Meeting on Interconnection of Molecular Biology Databases (MIMBD-95) was held at Cambridge University on July 20-22, 1995, in conjunction with ISMB-95. The workshop was organized by Peter Karp of the SRI International Artificial Intelligence Center, Victor M. Markowitz of Lawrence Berkeley National Laboratory, and Tom Flores of the European Bioinformatics Institute. The premise of this meeting was that the roughly 100 existing molecular biology databases (MBDs) will be of much greater value to molecular biologists when interconnected than in their current isolated states. Scientists will be able to integrate diverse sources of information to answer questions that are laborious or impossible to tackle today.

Research in this area is proceeding along three main lines: interconnecting databases via WWW, the data-warehousing approach, and the distributed-query approach. Central to all three approaches is the concept of creating dfatabase links recording information about relationships between objects in various MBDs (such as linking a protein sequence in one to the corresponding DNA sequence in another).

The WWW-based approach to interoperation combines WWW access to entries of individual MBDs with electronic links to allow user navigation among entries of related databases. The distributed-query approach allows the expression of complex queries over multiple MBDs and addresses the problems of splitting such queries into component subqueries for individual MBDs and assembling the results. The data-warehousing approach also supports the expression of complex queries over multiple databases. Such queries are simplified by physically integrating multiple MBDs into a centralized data warehouse.

Other topics discussed at the meeting included the automatic inference of links among MBDs, maintenance of link integrity as objects are updated, similarities of notations (data-definition languages) employed for specifying structure (schema), and development of new MBDs.

Two panels were held at the workshop. The first panel discussed the role of standards and componentry in achieving database interoperation and reuse of bioinformatics software. The second panel discussed the current state of systems developed for supporting interoperability, including their architectures, approaches, assumptions, limitations, and costs; and explored future directions. Position papers presented during the second panel were published in Journal of Computational Biology 2(4), 1995.

As a result of the growing maturity of MBD interoperability work, next year's MIMBD meeting will be merged into the ISMB-96 meeting in St. Louis. A one-day follow-up MIMBD workshop will also be held in conjunction with ISMB-96 to discuss advances in this area.