Protein-Protein Complex Docking Prediction


Protein-protein docking is a complex and important topic. From the initial rigid docking ZDOCK based on the FFTs algorithm to the multi-step integration of HADDOCK, ClusPro, SwamDock, etc., the algorithms in this field are continuously upgraded and iterated. RosettaDock is a veteran in the field of protein-protein docking, and has been tested by CAPRI for a long time. Particularly good at exploring the local conformation of protein-protein. Including many articles that have been published, RosettaDock is used for final optimization, and its status is evident. In recent years, RosettaDock has also developed algorithms for special protein families, such as SnugDock for antibody-antigen docking, SymmetricDock for homopolymer assembly docking, and FlexPepDock for peptide-protein docking, as well as those developed for difficult docking tasks. RosettaDock4.0 of Motif Score and so on. This article will briefly introduce the basic framework and usage of conventional RosettaDock3.2 protein-protein docking.

Basic principles and algorithms

The basic algorithm of RosettaDock is as follows:

The RosettaDock algorithm.Figure 1. The RosettaDock algorithm. (Chaudhury S, et al. 2011)

The entire docking process is divided into two parts. In the first low-resolution stage, the side chain conformation between the proteins is replaced by a coarse-grained sphere. Search directly for the degree of matching of the skeleton shape between the proteins. In the second stage, all side chain conformations will be considered to calculate more accurate interaction energy.

The initial local interference will randomly translate and rotate one of the components of the first guessed conformation by 8 angstroms and 8° (or 8 angstroms, 3°).

In the low-precision stage, 500 movements of the rigid body are performed, in which it is possible to choose whether to perform the conformational exchange selection of Ensemble. Output the lowest conformation for high-precision stage.

In the high-precision stage, perform MCMPCycle 50 times:

Repack conformation and minimize energy, as the initial starting conformation.

Internal cycle: Run MCMCycle 50 times, each MCMCycle includes: rigid body movement, RotamerTrials optimize each amino acid to the lowest energy state, and judge whether the energy is greater than 15REU, if the energy drop is too small, the energy between the rigid body is minimized once. (MCMCycle runs Repack every 8 steps (repack mode can be divided into rt_min or sc_min))

Outer loop: Reset the initial state and run the inner loop that is repeated 4 times

Finally, the lowest energy conformation is restored from the entire mc trajectory (the energy minimization will be performed again at the end).

Basic tutorial of RosettaDock

3.1 Preliminary structure preparation

RosettaDock is a docking algorithm based on MCM and is different from other rigid docking software. It is not good at global docking because the calculation efficiency of the overall process is too low. Therefore, before RosettaDock, ZDOCK etc. are generally used for preliminary conformation exploration. You can choose one or more seemingly reasonable conformations as a starting point.

Optional FFTs server:

  • ZDOCK server:
  • SmoothDock server:
  • ClusPro:
  • Haddock:

3.2 Run RosettaDock

It is very convenient to run RosettaDock, one command is enough. Run the local_dock standard calculation process:
docking_protocol.linuxgccrelease -s prepacked.pdb -partners H_A -dock_pert 3 8 -ex1 -ex2aro

3.3 Local conformation optimization

Sometimes, the final conformation we get may be due to some small skeletal conflicts resulting in higher overall energy of the complex, then we can adopt a less aggressive sampling strategy at this time.

docking_protocol.linuxgccrelease -s prepacked.pdb -partners H_A -dock_pert 3 8 -ex1 -ex2aro -spin -use_input_sc -dock_mcm_trans_magnitude 0.7 -dock_mcm_rot_magnitude 5.0 -nstruct 1 -docking_local_refine

4 Result analysis

4.1 Energy and interface parameter analysis

The official tutorial for protein-protein docking analysis recommends the use of total_score and I_sc. For preliminary analysis, it is completely sufficient. If you want to do more in-depth analysis, you can use the Rosetta InterfaceAnalyzer app to perform complex interface analysis.

4.2 Preliminary guess of the influence of conformation

In addition, the accuracy and credibility of protein-protein docking have a relatively large relationship with the initial conformation. If the initial guessed conformation is far from the real, the optimized conformation will also deviate farther.

5 How to better improve the accuracy of the model?

Data-driven HADDCOKing.Figure 2. Data-driven HADDCOKing. (Dominguel, Boelens & Bonvin. 2003)

In RosettaDock, we can use limiting parameters to influence the docking process. Here are some strategies for integrating and docking available experimental information:

  • Use experimental point mutation information to find the basic binding area and limit the distance between hot residues.
  • If there is a complex of homologous proteins, we can also directly align the docking components to the homologous protein as a preliminary guess of the conformation (great help).
  • With the reduction of the cost of mass spectrometry cross-linking technology, information about the interaction pairs of amino acids at the interface of the complex can be obtained, helping to generate initial guessed conformations and generate distance limits.
  • According to the literature, inactive areas are blocked.


  1. Chaudhury S, Berrondo M, Weitzner B D, et al. Benchmarking and Analysis of Protein Docking Performance in Rosetta v3.2. Plos One, 2011, 6(8): e22477.
* For Research Use Only.