Figure 2. Structural modeling of protein interactome.
(a) ~50% of protein complexes with interfaces modeled by high-throughput techniques have accuracy suitable for docking. (b) Models by template-based docking classified according to CAPRI-like criteria (clashes and contacts not considered). High accuracy: ligand (the smaller protein in the complex) RMSD < 1.0 Å or interface RMSD (measured on the interface residues Cα) < 1.0 Å. Medium accuracy: ligand RMSD < 5.0 Å or interface RMSD < 2.0 Å. Acceptable accuracy: ligand RMSD < 10.0 Å or interface RMSD < 4.0 Å. The data indicates that even inaccurate models typically dock with good success rate. (c) Availability of protein-protein structures in genomes with the largest number of known protein interactions (according to BIND [227] and DIP [228] databases). “No model” indicates no template for the interactors, not the complex (almost all structurally characterized interactors have templates for their complexes).