Most genes, proteins and other components carry out their functions within a complex network of interactions and a single molecule can affect a wide range of other cell components. A global, integrative, approach has been developed for several years, including protein-protein interaction networks (interactomes). In this review, we describe the high-throughput methods used to identify new interactions and to build large interaction datasets. The minimum information required for reporting a molecular interaction experiment (MIMIx) has been defined as a standard for storing data in publicly available interaction databases. Several examples of interaction networks from molecular machines (proteasome) or organelles (phagosome, mitochondrion) to whole organisms (viruses, bacteria, yeast, fly, and worm) are given and attempts to cover the entire human interaction network are discussed. The methods used to perform the topological analysis of interaction networks and to extract biological information from them are presented. These investigations have provided clues on protein functions, signalling and metabolic pathways, and physiological processes, unraveled the molecular basis of some diseases (cancer, infectious diseases), and will be very useful to identify new therapeutic targets and for drug discovery. A major challenge is now to integrate data from different sources (interactome, transcriptome, phenome, localization) to switch from static to dynamic interaction networks. The merging of a viral interactome and the human interactome has been used to simulate viral infection, paving the way for future studies aiming at providing molecular basis of human diseases.