Transparent access to multiple bioinformatics information sources

This paper describes the Transparent Access to Multiple Bioinformatics Info rmation Sources project, known as TAMBIS, in which a domain ontology for mo lecular biology and bioinformatics is used in a retrieval-based information integration system for biologists. The ontology, represented using a descr iption logic and managed by a terminology server, is used both to drive a v isual query interface and as a global schema against which complex intersou rce queries are expressed. These source-independent declarative queries are then rewritten into collections of ordered source-dependent queries for ex ecution by a middleware layer. In bioinformatics, the majority of data sour ces are not databases but tools with limited accessible interfaces. The ont ology helps manage the interoperation between these resources. The paper em phasizes the central role that is played by the ontology in the system. The project distinguishes itself from others in the following ways: the ontolo gy, developed by a biologist, is substantial; the retrieval interface is so phisticated; the description logic is managed by a sophisticated terminolog y server. A full pilot application is available as a Java (TM) applet integ rating five sources concerned with proteins. This pilot is currently underg oing field trials with working biologists and is being used to answer real questions in biology, one of which is used as a case study throughout the p aper.