Treffer: PROVA: Rule-based java-scripting for a bioinformatics semantic web

Transparent information integration across distributed and heterogeneous data sources and computational tools is a prime concern for bioinformatics. Recently, there have been proposals for a semantic web addressing these requirements. A promising approach for such a semantic web are the integration of rules to specify and implement workflows and object-orientation to cater for computational aspects. We present PROVA, a Java-based rule-engine, which realises this integration. It enables one to separate a declarative description of information workflows from any implementation details and thus easily create and maintain code. We show how PROVA is used to compose an information and computation workflow involving - rules for specifying the workflow, - rules for reasoning over the data, - rules for accessing flat files, databases, and other services, and - rules involving heavy-duty computations. The resulting code is very compact and re-usable. We give a detailed account of PROVA and document its use with a example of a system, PSIMAP, which derives domain-domain interactions from multidomain structures in the PDB using the SCOP domain and superfamily definitions. PSIMAP is a typical bioinformatics application in that it integrates disparate information resources in different formats (flat files (PDB) and database (SCOP)) requiring additional computations. PROVA is available at comas.soi.city.ac.uk/prova.