Many application domains require data stored in heterogeneous SQL and NoSQL data sources. Improving applications and end-user access to such heterogeneous data is now more important than ever. Our work contributes to this objective by facilitating complex data exploration. We adopt a graph data model to depict, in an integrated view, the data available in the sources and to represent a conceptual schema of that data. Our purpose is to provide a conceptual query language to facilitate end-users (e.g. medical domain experts), retrieving data from those sources through ad hoc queries. As part of this study, in this paper we propose a set of high-level operators to query data by transforming graphs. These operators allow successive graph transformations to facilitate generation of graphs with filtered data, and with new relationships deriving information that is implicit or that is sparse from the data.