Résumé: In silico scientific experiments encompass multiple combinations of program and data resources. Each resource combination in an execution flow is called a scientific workflow. In bioinformatics environments, program composition is a frequent operation, requiring complex management. A scientist faces many challenges when building an experiment: finding the right program to use, the adequate parameters to tune, managing input/ output data, building and reusing work flows. Typically, these workflows are implemented using script languages because of their simplicity, despite their speci ficity and di .culty of reuse. In contrast, Web service technology was specially conceived to encapsulate and combine programs and data, providing interoperation between applications from different platforms. The Web services approach is superior to scripts with regard to interoperability, scalability and flexibility issues. We have combined metadata support with Web services within a framework that supports scientific workflows. While most works are focused on metadata issues to manage and integrate heterogeneous scienti .c data sources, in this work we concentrate on metadata support to program management within workflows. We have used this framework with a real structural genomic workflow, showing its viability and evidencing its advantages.

Langue: Anglais