Polar residue hot spots have been observed at protein-protein binding sites. Here we show that hot spots occur predominantly at the interfaces of macromolecular complexes, distinguishing binding sites from the remainder of the surface. Consequently, hot spots can be used to define binding epitopes. We further show a correspondence between energy hot spots and structurally conserved residues. The number of structurally conserved residues, particularly of high ranking energy hot spots, increases with the binding site contact size. This finding may suggest that effectively dispersing hot spots within a large contact area, rather than compactly clustering them, may be a strategy to sustain essential key interactions while still allowing certain protein flexibility at the interface. Thus, most conserved polar residues at the binding interfaces confer rigidity to minimize the entropic cost on binding, whereas surrounding residues form a flexible cushion. Furthermore, our finding that similar residue hot spots occur across different protein families suggests that affinity and specificity are not necessarily coupled: higher affinity does not directly imply greater specificity. Conservation of Trp on the protein surface indicates a highly likely binding site. To a lesser extent, conservation of Phe and Met also imply a binding site. For all three residues, there is a significant conservation in binding sites, whereas there is no conservation on the exposed surface. A hybrid strategy, mapping sequence alignment onto a single structure illustrates the possibility of binding site identification around these three residues.protein-protein interfaces ͉ hot spots ͉ molecular recognition ͉ binding site prediction ͉ residue conservation R inge (1) has raised the question ''what makes a binding site a binding site?'' Many studies have addressed this intriguing and vastly important problem. Being able to a priori predict binding sites would both limit the conformational search in drug design, facilitate the prediction of protein-protein interactions (2), and may provide leads to binding site design.A number of studies have examined the attributes of proteinbinding sites (3-5). Although binding sites on enzyme surfaces typically consist of a concave cleft shape (6, 7) and similarly small ligand binding sites on receptor surfaces (8), this is not the case for the larger protein-protein complexes (9-12). Enzyme-binding sites were shown to frequently be the largest cavities on the enzyme surface (6, 7). On the other hand, the shape of dimer-binding sites is usually quite flat (9) and practically indistinguishable from other patches on the protein surface. Native binding sites do not yield the largest possible interfaces between two protein molecules. A docking study has shown that nonnative interfaces can be larger, and bury a larger extent of total or nonpolar surface areas (13). A similar observation has been made for the number of salt bridges or hydrogen bonds (13,14). Hence, although interfaces are frequently largely hydrophobic ...