Agent interaction for bioinformatics data management

As genome projects produce increasingly large quantities of sequence data, fast and reliable sequence analysis methods are required. Basic methods for comparing pairs of sequences or detecting patterns are well-developed, and now the key problem in analyzing this genomic data is how to integrate the software and primary databases in a flexible and robust way. The wide rang e of available programs conform to very different input, output, and proces sing requirements, typically with little consideration given to issues of i ntegration. Key to addressing these issues appropriately is not to consider them as a result of the biological domain, but instead as an information p rocessing problem that suggests nothing as much as an agent-based approach. In this paper, we introduce GeneWeaver, a multi-agent system for bio-infor matics, and describe in detail the agent interactions which allow the integ ration and management of analysis methods and data. The system does not off er new methods but instead manages existing databases and analysis tools in an effective and flexible way, and facilitates easy and dynamic growth.