Complex applications require integration systems that access data locally or on the Web while being capable to exploit and complete the various capabilities available at integrated resources. In scientific applications such as biomedical, engineering or geographical, information systems (IS) are highly heterogeneous: they differ by their data representation and by their radically different query languages. Therefore, in addition to the common problem of data integration, provided scientific query languages shall also be integrated. In this paper we propose an approach that not only focuses on the data integration, but also addresses the integration of query capabilities available at the sources. An IS may provide a query capability inexistent at another IS, whereas two query capabilities may be similar but with slightly different semantics. We introduce the notion of derived wrapper that captures additional query capabilities to either compensate capabilities lacking at a source, or to adjust an existing capability in order to make it homogeneous with other similar capabilities, wrapped at other sources. The use of derived wrappers extends traditional mediation approaches.