With the growing tools of small molecules, the modulation of these networks of interactions represents 
a promising therapeutic strategy. Protein-protein interaction inhibitors are certainly the next generation of highly innovative drugs that will reach the market in the next decade. As a consequence of this enthusiasm, the exponential increase of published biomedical literature on PPIs and their inhibition has prompted the development of internet services and databases that help scientists to manage the available information. There is now a growing number of structural databases dedicated to proteinprotein interactions. A large variety of these PPIs databases depict protein-protein interactions at a structural level, but they focus only on this particular interface without taking into account the potential inhibitors related to one of the two partners. In a recent survey, Higuerueolo et al. analyzed the atomic interactions and profile of small molecules disrupting PPIs in the TIMBAL database, focusing on small molecules properties and comparing these results to drug-like databases. Several other studies have also focused on subsets of small molecules that disrupt PPIs. However, none of them have focused on both the protein-protein structural information available and the known inhibitors within the interface. We describe here a chemical space, 2P2IDB, which is a handcurated database dedicated to the structure of Protein-Protein complexes with known inhibitors thereby offering complementary information to these previous analyses. We have analyzed the protein/protein and protein/inhibitor interfaces in terms of geometrical parameters, atom and residue properties, buried accessible surface area and other biophysical parameters, such as the protein-protein dissociation constant of a complex. The interfaces found in 2P2IDB were then compared to those of representative datasets of heterodimeric complexes from Bahadur and Zacharias or from the ProtorP parameters of a globular protein, or a short peptide. Complexes in 2P2IDB present globally the same shape than standard heterodimeric complexes, but their accessible surface areas are significantly smaller. More strikingly, no major conformational changes are observed between the different states of the proteins. The interfaces are also more hydrophobic than general PPIs�� interfaces, with less charged residues and more non-polar atoms. Moreover, fifty percent of the complexes in the 2P2IDB dataset possess more hydrogen bonds than typical protein-protein complexes. A set of key descriptors were identified to distinguish between PPIs with known inhibitors and representative transient complexes in the protein databank. Transient protein-protein complexes are defined as protomers that, in vivo, can exist either on their own or in complex and also undergo an exchange between the free and complexed form. A new classification based on these parameters is proposed with potential aims for the future to identify potential new druggable PPI targets. As our goal was to define structural parameters that guide the development of PPI disruptors, we only considered those protein families for which a high resolution three dimensional structure was available for both the protein/protein and the protein/ inhibitor complexes. SCH772984 942183-80-4 Homodimers and covalently bound inhibitors were not taken into account due to their different behavior. When available, the best resolution structure of the LY2835219 in vivo unbound form of the proteins or a close homologue was included. The dataset was built through data mining from the literature and by exhaustive search of the Protein Data Bank.