Despite similar protein folds, as few as of residues within similar protein 3D structures can form a common core. RMS deviations on core atoms can be as high as Å. Similar protein structures can have secondary structure identities as low as , which is equivalent to that expected by chance. By defining three categories of amino acid accessibility (buried, half buried and exposed), some similar protein 3D structures have as few as of positions in the same category, making them indistinguishable from pairs of dissimilar protein structures. Similar structures can also have as few as of common side-chain to side-chain contacts, and virtually no similar energetically favourable side-chain to side-chain interactions. Complementary changes are defined as structurally equivalent pairs of interacting residues in two structures with energetically favourable but different side-chain interactions. For many proteins with similar three-dimensional structures, the proprotion of complementary changes is near to that expected by chance, suggesting that many similar structures have fundamentally different stabilising interactions.

All of the results suggest that proteins having similar 3D structures can have little in common apart from a scaffold of common core secondary structures. This has profound implications for methods of protein fold detection, since many of the properties assumed to be conserved across similar protein 3D structures (e.g., accessibility, side-chain to side-chain contacts, etc.) are often unconserved within weakly similar (i.e., type and ) protein 3D structures. Little difference was found between type and similarities suggesting that the structure of similar proteins can evolve beyond recognition even when function is conserved.

Our findings suggest that it is more general features of protein structure, such as the requirements for burial of hydrophobic residues and exposure of polar residues, rather than specific residue-residue interactions that determine how well a particular sequence adopts a particular fold. If detection of similar folds having little in common outside of their core secondary structures is to become a reality, efforts should concentrate on such general principles, and on methods for modelling large loop regions that are likely to differ between similar 3D structures.