The receptor tyrosine kinase c-Met have multiple roles during cancer development and is currently considered as an important target for molecularly targeted therapies. Structural knowledge of how compounds interact on c-Met catalytic site could guide structure-based drug design strategies towards more effective and selective anticancer drug candidates. However, although 17 crystal structures of c-Met complexed with adenosine triphosphate (ATP)-competitive kinase inhibitors are publicly available (August 2009), there are still open questions regarding the prediction of ligand binding modes. We have applied molecular modeling and molecular mechanics to analyze the distribution of ligands interaction energy on c-Met residues, and deduced a new model of the active site allowing for an unambiguous identification of ligand binding modes. We demonstrate that the binding of known ligands on the c-Met catalytic site involves seven identified structurally-distinct areas. Five of these match the generic kinase ATP binding site model built by Novartis scientists in the 1990s, while the two others are distinct allosteric regions that can be exploited by second generation kinase inhibitors such as Gleevec. We show here that c-Met can accept both such kinds of allosteric inhibitors, a very unusual feature in the kinase family that opens new grounds for highly specific drug design.