Assessment and improvement of the reliability of protein-protein interaction (ppi) data is critical for the progress of the currently active research on interactomes. Some interesting questions in this respect are: How three-dimensional (3D) protein structural data is present in known ppi data?, and How this kind of information can be used to validate and improve the interactomes? To address this problem, analysis and unification of six structural domain-domain interaction (sddi) datasets is presented; followed by a comparative study of these sddi data in three ppi reference sets produced at different levels of confidence. The results show that protein structural and interactomic data are partially complementary and that a larger proportion of structural information is observed in more confident interactomes. We also present, focused on the human interactome, an analysis of the domains that are more frequently present in: (i) an interactome based on validation by at least two experimental methods versus (ii) another interactome based on validation by 3D structural interaction data. These results allow to distinguish between domain pairs associated to protein interactions supported by 3D structures and domain pairs that at present are not supported by structural information. The domain pairs exclusive of interactions without associated 3D data reveal interacting conserved modules that are probably flexible, disordered, and difficult to crystallize; and which are often found in proteins involved in signaling pathways and DNA processing.