Dockground: A comprehensive data resource for modeling of protein complexes

Abstract

Characterization of life processes at the molecular level requires structural details of protein interactions. The number of experimentally determined structures of protein-protein complexes accounts only for a fraction of known protein interactions. This gap in structural description of the interactome has to be bridged by modeling. An essential part of the development of structural modeling/docking techniques for protein interactions is databases of protein-protein complexes. They are necessary for studying protein interfaces, providing a knowledge base for docking algorithms, and developing intermolecular potentials, search procedures, and scoring functions. Development of protein-protein docking techniques requires thorough benchmarking of different parts of the docking protocols on carefully curated sets of protein-protein complexes. We present a comprehensive description of the Dockground resource (http://dockground.compbio.ku.edu) for structural modeling of protein interactions, including previously unpublished unbound docking benchmark set 4, and the X-ray docking decoy set 2. The resource offers a variety of interconnected datasets of protein-protein complexes and other data for the development and testing of different aspects of protein docking methodologies. Based on protein-protein complexes extracted from the PDB biounit files, Dockground offers sets of X-ray unbound, simulated unbound, model, and docking decoy structures. All datasets are freely available for download, as a whole or selecting specific structures, through a user-friendly interface on one integrated website.

Keywords: benchmark sets; protein recognition; protein-protein interactions; structure prediction.

Figure 1

Schematic representation of interconnectivity between different Dockground modules and their relation to the external PDB. Stroked arrows represent future developments.

Figure 2

Dockground bound page, along with the Quick Downloads. The detailed search part is shown in Figure 3.

Figure 3

Dockground search screen and fragment of the corresponding search results.

Figure 4

Statistics on the basic (bound) part of Dockground. Normalization of data is with respect to the total number of PDB biounit files (A, B), the total number of chains (C), and the total number of pairwise complexes (C, D) in the database.

Figure 5

Comparison of complexes in the Weng's benchmark set 5 and the Dockground set 4. (A) Docking difficulty level (adopted from Ref. 19). Rigid‐body (or easy) cases correspond to the protein complexes with bound/unbound interface C^α RMSD (i‐RMSD) ≤ 1.5 Å; medium difficulty cases correspond to 1.5 < i‐RMSD < 2.2 Å; and difficult cases correspond to i‐RMSD >2.2 Å. (B) Functional categories of complexes, as in Ref. 19: EI, Enzyme‐Inhibitor; ES, Enzyme‐Substrate; ER, Enzyme complex with a regulatory or accessory chain; A, Antibody‐Antigen; AB, Antigen‐Bound Antibody; OG, Other, G‐protein containing; OR, Other, Receptor containing; and OX, Other, miscellaneous. The figure shows the absolute number of complexes in each category.

Figure 6

Example of docking decoys. False positive matches are shown by the ligand (smaller protein) center of mass (red) for 1f93 complex. The near‐native match is indicated by an arrow. The native structure of the complex is in cartoon representation.