Distributions of optimality for affinity [Equation (1)] and specificity [Equation (3)] in four catagories of protein–DNA interfaces. Red: helical transcription factors; green: restriction endonucleases; blue: homing endonucleases; black: nonspecific enzymes. Histogram bin centers are indicated on the horizontal axes. Only positions at which mutation to glycine is predicted to result in the loss of >3 kcal/mol of binding energy were included.

Optimality of affinity and specificity at individual positions in representative complexes. Protein identities and pdb codes are indicated at the top of each panel. In the left panels, the extent optimal for affinity [Equation (1)] is plotted against the extent optimal for specificity [Equation (3)] for each residue in a complex. The color of each data point is proportional to the change in the energy of binding calculated for mutation to glycine at that position, where bright cyan indicates the highest relative loss in binding energy, and black indicates the lowest loss in binding energy. All interface residues in each crystal structure are represented; large clusters near (0, 0) correspond to residues in the interface that are not predicted to be optimal for affinity or specificity. At right are representations of the crystal structures for each indicated interface. The amino acids are colored by a dual color gradient in which red indicates the optimality for affinity, blue indicates the optimality for specificity and pink/magenta indicates positions that are optimal for both affinity and specificity. A) DNAse I (pdb code: 2DNJ), B) C2H2 zinc finger Zif268 (1ZAA), C) [β]-Zip GCN4 (2DGC), D) EcoRV (1B94), E) I-MsoI (1M5X). In frame (D), symmetrically equivalent residues in the homodimer EcoRV are labeled on only one chain. Molecular images were rendered using PyMOL (11).