Shared interactions

Figure 11 shows the shared interactions (i.e., the number of residue-residue interactions common to both structures divided by the smallest number of interactions in the pair of aligned structures) versus . Surprisingly, structurally similar proteins can have as few as of their interactions in common.

The above definition of interacting residues is somewhat strict in that it requires that there be at least one side-chain to side-chain contact between pairs of residues. Relaxing the requirement and defining interacting residues as those residues with atoms (or built in the case of glycine residues) within Å of each other gives the analogous plots in Figure 12. By this relaxed definition, the percentage of shared interactions tends to be higher, but structurally similar proteins can still have as few as of interactions in common. The greater separation between similar and dissimilar protein pairs suggests that this geometrically less exact consideration of side-chain interactions might be more effective as a tool for protein fold recognition.

Perhaps more interesting than shared interactions are shared favourable interactions. These are interactions common to both structures that both contribute a negative pseudo-energy term () Figure 13 shows the percentage of shared favourable interactions versus . As would be expected from Figure 10, proteins having highly similar sequences have about half of their interactions as shared and favourable ( is the approximate maximum). However, proteins having no detectable sequence similarity have less than of the total possible interactions as shared and favourable, and many distantly related structures have essentially no common favourable interactions. Many distantly related proteins have a proportion of shared favourable interactions near to that expected by chance (see dashed line in Figure 13 and Table 4), suggesting that many pairs of structurally similar proteins have completely different stabilising interactions.

A notable outlier in Figures 11 and 13 is the similarity between Ricin domain 2 (1AAIB_II) and Interleukin 1 (81IB). This pair of proteins has a shared interactions greater than others of a similar . This might be explained by the conservation of particular amino acids (and their corresponding side-chain to side-chain interactions) within the trefoil family of proteins [Murzin et al., 1992], though the fact that other trefoil pairs of a similar do not have such a high percentage of shared interactions might suggest that the Ricin/Interleukin-1- similarity is fortuitous.

Unlike secondary structure and accessibility, genuine structural similarities tend to have more common interactions than dissimilar structures. In Figures 11 and 12 there is a distinct separation between structurally similar and dissimilar proteins. This difference is less pronounced for favourable pairs (Figure 13).