Protein-protein interactions (PPIs) are essential, and modulating their function through PPI-targeted drugs is an important research field. PPI sites are shallow protein surfaces readily accessible to the solvent, thus lacking a proper pocket to fit a drug, while their lack of endogenous ligands prevents drug design by chemical similarity. The development of PPI-blocking compounds is, therefore, a tough challenge. Mixed solvent molecular dynamics has been shown to reveal protein-ligand interaction hot spots in protein active sites by identifying solvent sites (SSs). Furthermore, our group has shown that SSs significantly improve protein-ligand docking. In the present work, we extend our analysis to PPI sites. In particular, we analyzed water, ethanol, and phenol-derived sites in terms of their capacity to predict protein-drug and protein-protein interactions. Subsequently, we show how this information can be incorporated to improve both protein-ligand and protein-protein docking. Finally, we highlight the presence of aromatic clusters as key elements of the corresponding interactions.