Symmetric proteins play important roles in many biological processes, such as signal transduction and molecular transportation. Therefore, determining the symmetric oligomeric structure of subunits is crucial to investigate the molecular mechanism of the related processes. Due to the high cost and technical difficulties associated with many experimental methods, computational approaches, such as molecular docking, have played an important complementary role in the determination of symmetric complex structures, in which a benchmark data set is pressingly needed. In the present work, we develop a comprehensive and non-redundant benchmark for symmetric protein docking based on the structures in the Protein Data Bank (PDB). The diverse dataset consists of 251 targets, including 212 cases with cyclic groups symmetry, 35 cases with dihedral groups symmetry, 3 cases with cubic groups symmetry, and 1 case with helical symmetry. According to the conformational changes in the interface between bound and unbound structures, the 251 targets were classified into three groups: 176 "easy", 37 "medium", and 38 "difficult" cases. A preliminary docking test on the targets of cyclic groups symmetry with M-ZDOCK indicated that symmetric multimer docking remains challenging. The benchmark will be beneficial for the development of symmetric protein docking algorithms. The proposed benchmark data set is available for download at http://huanglab.phys.hust.edu.cn/SDBenchmark/.

Protein–protein interactions are crucial in many biological processes. Therefore, determining the complex structure between proteins is valuable for understanding the molecular mechanism and developing drugs. Many proteins like ion channels are formed by symmetric homo-oligomers. In this study, we have proposed a hierarchical docking algorithm to predict the structure of C _n symmetric protein complexes, which is referred to as CHDOCK. The symmetric binding modes were first constructed by an FFT-based docking algorithm and then optimized by our iterative scoring function for protein–protein interactions. When tested on a symmetric protein docking benchmark of 212 homo-oligomeric complexes with C _n symmetry, CHDOCK obtained a significantly better performance in binding mode predictions than three state-of-the-art symmetric docking methods, M-ZDOCK, SAM, and SymmDock. When the top 10 predictions were considered, CHDOCK achieved a success rate of 44.81% and 72.17% for unbound docking and bound docking, respectively in comparison to those of 36.79% and 65.09% for M-ZDOCK, 31.60% and 54.25% for SAM, and 30.66% and 31.60% for SymmDock. CHDOCK is computationally efficient and can normally complete a symmetric docking calculation within 30 min. The CHDOCK can be freely accessed by a web server athttp://huanglab.phys.hust.edu.cn/hsymdock/.