Theoretical and experimental restraints to drive the docking of protein-protein complexes

Biological processes are frequently driven by protein-protein interactions. The number of known protein interactions is much higher than the number of known protein complex structures. To bridge this gap, data-driven protein-protein docking utilizing experimental or theoretical restraints is applied. In this study the PROTIN_ID method for generating theoretical docking restraints is introduced. PROTIN_ID generates residue clusters on the protein surface based on sequence conservation. Compared to WHISCY and CCRXP, PRO TIN_ ID performs equally well or better. Furthermore, PROTIN_ID has user-friendly features such as the ability to improve the quality of sequence alignments, which improves its performance, and automatically utilizing up-to-date sequence data for experimentally determined proteins or homology models to generate theoretical restraints. A webserver version of PROTIN_ID was implemented for the academic community. Statistical analyses of the conservation of interface residues using the latest version of Benchmark4.0 demonstrated that interface residues are more conserved than non-interface residues. The application of spatial clustering of residues is more efficient to exploit the conservation signal of interface residues, resulting in reliable predictions that are better than predictions generated by 'non-clustering' or at random. Theoretical restraints derived from PROTIN_ID were applied to drive docking and compared to ab initio docking, demonstrating that data-driven docking was more successful. Combining theoretical and experimental restraints to drive docking was compared to experimental-data driven docking. It was shown that combined restraints-driven docking improved because of increased interface residue recall, demonstrating that consensus-data is possibly useful for improvement of docking performance.