Only 24% of models clustered

The HADDOCK category is meant to discuss any HADDOCK-related issue. For general information about HADDOCK refer to

Dear HADDOCK team,

My docking result have given me 8 clusters while there are not many structures in each cluster(from 4 to 16). I noticed there’re only 24% of the water refined models were clustered. Please see below

“HADDOCK clustered 49 structures in 8 cluster(s), which represents 24 % of the water-refined models HADDOCK generated.”

I’m not sure how should I understand this result that I have many structures can’t be clustered. Is this normal? What is this result suggesting? ( I have used the default FCC clustering parameters)

Please let me know your suggestions, I’m sorry if I missed a similar question.

It all depends on how much information you defined to guide your docking.

I have only provided two unambiguous restraints to guide the docking. Does this mean that the structures and the restraints are not very compatible to form a complex model?

There are many things that can be at the origin of a poor cluster. It all depends on your input data, and parameter settings:

  1. Did you submit an ensemble of models?
  2. What kind of molecule are you docking
  3. Did you define full flexibility?
  4. Did you change the clustering settings?

As for your restraints, you should check in the output file of one it0 model, if they have been read properly?
Check also if they are satisfied.

You can also re-cluster manually with different settings. And visually inspect the models

Sometimes reaching a proper binding mode might be hard (e.g. partially buried sites) and the convergence poor.
This does not mean per se that you results are bad.

More info:

Best practice guide: – Bonvin Lab
Online manual - analysis: HADDOCK2.4 manual - Analysis – Bonvin Lab

Thank you very much for the suggestions. The Dos and Don’ts in the practice guide are also quite useful.

I compared the structures in the top4 clusters (their HADDOCK scores are not quite separated), looks like the two proteins were docked in four largely different positions and they all good for the 2 restraints I provided. I think my issue here could be just the restraints I have were not enough to get a good convergence. Those clusters might just all be possible solutions.