A Global and Local Structure-Based Method for Predicting Binary Protein-Protein Interaction Partners: Proof of Principle and Feasibility

We report a 3D structure-based method of predicting protein-protein interaction partners. It involves screening for pairs of tetrahedra representing interacting amino acids at the interface of the protein-protein complex, with one tetrahedron on each protomer. H-bonds and VDW interactions at their interface are first determined and then interacting tetrahedral motifs (one from each protomer) representing backbone or side chain centroids of the interacting amino acids, are then built. The method requires that the protein protomers be transformed first into double-centroid reduced representation (Reyes, V.M. & Sheth, V.N., 2011; Reyes, V.M., 2015a). The method is applied to a set of 801 protein structures in the PDB with unknown functions, which were screened for pairs of tetrahedral motifs characteristic of nine binary complexes, namely: (1.) RAP-Gmppnp-cRAF1 Ras-binding domain; (2.) RHOA-protein kinase PKN/PRK1 effector domain; (3.) RAC-HOGD1; (4.) RAC-P67PHOX; (5.) kinase-associated phosphatase (KAP)-phosphoCDK2; (6.) Ig Fc-protein A fragment B; (7.) Ig light chain dimers; (8.) beta catenin-HTCF-4; and (9.) IL-2 homodimers. Our search method found 33, 297, 62, 63, 120, 0, 108, 16 and 504 putative complexes, respectively. After considering the degree of interface overlap between the protomers, these numbers were significantly trimmed down to 4, 2, 1, 8, 3, 0, 1, 1 and 1, respectively. Negative and positive control experiments indicate that the screening process has acceptable specificity and sensitivity. The results were further validated by applying the CP and TS methods (Reyes, V.M., 2015b) for the quantitative determination of interface burial and inter-protomer overlap in the complex. Our method is simple, fast and scalable, and once the partner interface 3D SMs are identified, they can be used to computationally dock the two protomers together to form the complex.