Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
ClusPro is a free web server for protein-protein docking. You give it two protein structures (a receptor and a ligand), it samples orientations of one against the other, groups the best-scoring poses into clusters and returns models ranked by cluster size. The server’s own page states that it should only be used for noncommercial purposes. This guide covers how to get access, how to set up a first job, how to read the output, and what to do when a job fails.
Reader task: a researcher who has two protein structures and needs a predicted complex, and wants to know exactly what to prepare, click and check. Routes and wording below were read from the ClusPro help and tutorial pages on 2026-11-07.
What ClusPro is used for in research
The ClusPro 2.0 home page describes the service as “protein-protein docking”. Docking here means predicting how two separate protein structures fit together as a complex. The server is run by research groups linked on its home page to the Vajda Lab and the ABC Group, with Boston University and Stony Brook University listed. Its tutorials cover several docking modes: beginning (default) docking, antibody, multimer, SAXS, restraints and peptide. For small-molecule ligands and drug-like compounds, a different class of tool applies; see molecular docking software for a worked AutoDock Vina walkthrough.
How to install or access it
ClusPro is a web server, so nothing is installed. Use the official site at cluspro.bu.edu. The home page offers use without an account, log in, sign up, reset password and retrieve username. The help page lists Sign Up and Login, and says bulk submissions should be arranged by contacting the team. The home page also states that all passwords were reset at the time it was read and directs users to the Reset Password link.
Licensing: the home and help pages state that ClusPro should only be used for noncommercial purposes. Check that your use fits before submitting unpublished or proprietary structures.
Step-by-step: a first docking job
These steps follow the “ClusPro 2.0 Beginning Docking Tutorial” on the server.
- Sign up and log in.
- Enter a job name. It is optional; if left blank, the job ID is used.
- Choose the receptor, either from the PDB or with the “Upload PDB” option.
- Optionally list chains for the receptor: single letters separated by white space. If none are given, all chains are used.
- Choose the ligand and, optionally, its chains the same way.
- Submit. The tutorial’s example uses receptor 1QQU chain A and ligand 1BA7 chain B, taken from Protein-Protein Docking Benchmark 3.
- Watch the queue page, which shows job status and how many jobs are ahead of yours.
- Wait for the completion email. The tutorial says jobs usually finish within a few hours, depending on queue length and protein size.
- Click the job ID to open results.
Pre-flight input checklist
- Non-standard residues. The help FAQ says an “unknown residues” error means non-standard residues are present in ATOM records. Remove them, or change those records to HETATM.
- Chains. Name only the chains that take part in the interface you care about; leaving the field empty uses all chains.
- Mode. Pick the mode that matches your data (table below) instead of leaving every job on the default.
Which scoring option or mode to choose
| Situation | What the ClusPro help page says |
|---|---|
| Dominant interaction forces are unknown | Use the balanced coefficient set. |
| Antibody-antigen complex | Use the antibody mode. |
| Other scoring variants | An electrostatic-favored variant is offered alongside balanced, and results can be viewed per coefficient set. |
| Extra experimental information | Dedicated tutorials exist for Multimer, SAXS, Restraints and Peptide modes. |
Output and how to read it
- Ranking. The top 10 models are shown by default; a dropdown shows more. The help page says to rank models by cluster size, the server’s own ordering, and advises against ranking by PIPER scores, which are provided but not designed for that purpose.
- Coefficient sets. The results page has views for electrostatic-favored and hydrophobic-favored sets. Click the number above a model image to download its PDB file, or download all displayed models or all models for all coefficient sets.
- Files. Each output file holds the receptor and ligand together. Open it in a viewer that supports multiple entries, such as PyMOL, or split the file. In the tutorial’s PyMOL example the receptor is named
model.000.00_0001and the ligandmodel.000.00_0002; the receptor stays fixed during docking, so the ligand is the molecule that moves between models.
Troubleshooting
- “Unknown residues” error: see the checklist above (ATOM versus HETATM records).
- Job still queued: the queue page shows the number of jobs ahead; completion is by email.
- Many jobs to run: the help page says to contact the team about bulk submissions rather than submitting them one by one.
- Unsure which mode fits: follow the table above, and use the mode-specific tutorial on the server.
Free alternatives
| Tool | What its own pages say |
|---|---|
| ClusPro | Protein-protein docking server; noncommercial use only. |
| HADDOCK web server (Utrecht University, Computational Structural Biology group) | Described as an information-driven flexible docking approach for biomolecular complexes, covering protein-protein, protein-nucleic acid and protein-ligand complexes and assemblies of more than two molecules; access is stated as free for non-profit users and requires registration. Restraints from interfaces, cross-links, NMR or cryo-EM data can be supplied. |
Read each server’s terms before use, and record which server and version produced a model so the analysis can be reproduced; see computational reproducibility and the FAIR4RS principles for research software. Deposit final coordinates in a data repository and cite software according to the software citation principles.
How to cite ClusPro
The ClusPro Papers page lists these core citations (read 2026-11-07; author lists abbreviated):
- Kozakov D, et al. (2017). The ClusPro web server for protein-protein docking. Nature Protocols, 12(2), 255-278.
- Vajda S, et al. (2017). New additions to the ClusPro server motivated by CAPRI. Proteins, 85(3), 435-444.
- Desta IT, et al. (2020). Performance and Its Limits in Rigid Body Protein-Protein Docking. Structure, 28(9), 1071-1081.
Mode-specific papers (peptide, antibody, restraints, SAXS, dimer classification) are listed on the same page and should be cited when those modes are used.
FAQ
What is ClusPro?
ClusPro 2.0 is a protein-protein docking web server hosted at cluspro.bu.edu, for noncommercial use.
Is ClusPro free?
The server pages state it should only be used for noncommercial purposes. Account creation is offered on the home page.
How long does a ClusPro job take?
The tutorial says usually within a few hours, depending on the queue and protein size, with an email on completion.
Which ClusPro model is best?
The help page says to rank by cluster size, not by PIPER scores.
Why do I get an “unknown residues” error?
Non-standard residues in ATOM records. Remove them or change them to HETATM.
Sources
- ClusPro 2.0 home page,
http://cluspro.bu.edu/home.php, read 2026-11-07. - ClusPro help page,
http://cluspro.bu.edu/help.php, read 2026-11-07. - ClusPro 2.0 Beginning Docking Tutorial,
http://cluspro.bu.edu/tut_dock.php, read 2026-11-07. - ClusPro publications page,
http://cluspro.bu.edu/publications.php, read 2026-11-07. - HADDOCK web server,
https://wenmr.science.uu.nl/haddock2.4, read 2026-11-07.







