Protein interface conservation across structure space

Abstract

With the advent of Systems Biology, the prediction of whether two proteins form a complex has become a problem of increased importance. A variety of experimental techniques have been applied to the problem, but three-dimensional structural information has not been widely exploited. Here we explore the range of applicability of such information by analyzing the extent to which the location of binding sites on protein surfaces is conserved among structural neighbors. We find, as expected, that interface conservation is most significant among proteins that have a clear evolutionary relationship, but that there is a significant level of conservation even among remote structural neighbors. This finding is consistent with recent evidence that information available from structural neighbors, independent of classification, should be exploited in the search for functional insights. The value of such structural information is highlighted through the development of a new protein interface prediction method, PredUs, that identifies what residues on protein surfaces are likely to participate in complexes with other proteins. The performance of PredUs, as measured through comparisons with other methods, suggests that relationships across protein structure space can be successfully exploited in the prediction of protein-protein interactions.

Fig. 1.

Types of geometric conservation and their measures. Protein complex A is compared here to three other complexes B, C, and D. Typically one subunit is superposed on a structurally similar subunit in the complex to which it is being compared (i.e., A1 would be superposed on B1) and the transformation that relates the first subunits is applied to the second so that all proteins are in the same coordinate system. Measures of conservation generally involve calculating: the transformation (translation/rotation) required to optimally superimpose the second subunits on each other (brown/green arrows); distances and angles between the centers of mass of the second subunit (brown/green spheres); and the alignment (independent of residue identity) of interfacial residues in a primary sequence alignment of the two subunits (red squares). Although there is some similarity between A and each of the other three complexes, recognizing it will depend on which measure is used (see text).

Fig. 2.

The surface of T-cell receptor protein CD8 (PDB ID code 1akj, chain D) colored according to the frequency with which interactions made by its structural neighbors are mapped to individual residues on its surface (red/white/blue = high/intermediate/low frequency). Each surface is colored based on a different set of structural neighbors: (A) SCOP family b.1.1.1; (B) superfamily b.1.1; (C) fold b.1; (D) PSD < 0.6 (found by Ska); (E) PSD < 0.6 in different families; (F) PSD < 0.6 in different superfamilies; (G) PSD < 0.6 in different folds. The red high contacting frequency regions show conserved protein interface.

Fig. 3.

Distributions of Z scores reflecting interface conservation. Each column in the graph shows a Z-score distribution when interface conservation for proteins in our docking benchmark set is calculated based on a different set of structural neighbors. The black bars and the width of each plot reflects the density of Z scores near the corresponding value on the y axis. Solid lines with green diamonds show the mean value of each distribution. The dashed line corresponds to a Z score of 3, which we take as the cutoff of statistical significance. The individual plots have been scaled so that their areas are proportional to the number of proteins for which a valid Z score could be calculated.

Fig. 4.

Calculating the contact map and contact frequency map. In the above example, a given query protein (Q, brown) with seven residues has five residues on the surface. Structural neighbors (Ni, green lines) involved in protein complexes are superimposed on Q, and the same transformation is applied to their interacting partners (Pi, green surfaces). Whenever a heavy atom from a residue of Pi is of an atom of a surface residue of Q after applying the transformation, that residue is marked (red circles), generating a contact map for each structural neighbor (black boxes represent nonsurface residues that are not included). The “contact frequency map” is generated by summing the individual contact maps.