Significance Protein−protein interfaces have become an emerging class of molecular targets for the design of therapeutic drugs. However, major challenges exist for the correct identification of binding sites on the protein surface as well as drug-like modulators of protein−protein interaction. An integrated approach using molecular fragment docking and coevolutionary analysis is presented to face these challenges. This approach can accurately predict and characterize the binding sites for protein−protein interactions as well as provide clusters of bound, fragment-sized molecules on the druggable regions of the predicted binding site. These bound, molecular fragments can be chemically combined to create candidate drugs.