# Haddock Z-score vs RMSD

**URL:** <https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185>\
**Category:** HADDOCK\
**Created:** [July 29, 2021, 1:41am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185 "2021-07-29T01:41:05Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![dinesh](https://avatars.discourse-cdn.com/v4/letter/d/a8b319/32.png) [@dinesh](https://ask.bioexcel.eu/u/dinesh)\
**Post date:** [July 29, 2021, 1:41am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185/1 "2021-07-29T01:41:05Z")

</div>

hi everyone.

I performed protein-protein docking with three sets of active residues. While analyzing the results of protein-protein docking, I am confused in choosing the best structure. Should I choose the structure from the cluster with low Z-score or should I choose a structure that gives the least RMSD which has a poor Z-score?

My aim is to identify the optimal set of active residues that can help reproduce the crystal structure.

---

<div class="post-metadata">

**Author:** ![amjjbonvin](https://dub1.discourse-cdn.com/flex013/user_avatar/ask.bioexcel.eu/amjjbonvin/32/23_2.png) [@amjjbonvin](https://ask.bioexcel.eu/u/amjjbonvin)\
**Post date:** [July 29, 2021, 6:45am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185/2 "2021-07-29T06:45:32Z")

</div>

RMSDs have no predictive values - they only tell where a cluster is with respect to the best model generated by HADDOCK

As we always indicate in all papers, it is the HADDOCK score which should be leading.

The z-score is only a statistical value related to a particular run.

And if you want to compare three runs with different definitions of active residues, it might be best to then remove the restraint energy term from the scores (subtract 10% of the restraint energy from the HADDOCK score).

---

<div class="post-metadata">

**Author:** ![dinesh](https://avatars.discourse-cdn.com/v4/letter/d/a8b319/32.png) [@dinesh](https://ask.bioexcel.eu/u/dinesh)\
**Post date:** [July 29, 2021, 6:52am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185/3 "2021-07-29T06:52:44Z")

</div>

thank you for your reply. But if I have to compare the resulted structures with an existing crystal structure, can I still use Haddock score to pick the best model?

---

<div class="post-metadata">

**Author:** ![amjjbonvin](https://dub1.discourse-cdn.com/flex013/user_avatar/ask.bioexcel.eu/amjjbonvin/32/23_2.png) [@amjjbonvin](https://ask.bioexcel.eu/u/amjjbonvin)\
**Post date:** [July 29, 2021, 10:41am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185/4 "2021-07-29T10:41:02Z")

</div>

In a case where you would not know the answer this would be the strategy.  
Although we always advise to look at more than one solution and try to validate those possibly with e.g. known mutations or any other info you might have (if not used to guide the docking).

But if you have a crystal structure you can evaluate all clusters

See our best practice guide for more tips:

> **[– Bonvin Lab](https://www.bonvinlab.org/software/bpg/#after-docking-how-to-analyse-docking-results-from-haddock-or-refine-models)**

---

<div class="post-metadata">

**Author:** ![dinesh](https://avatars.discourse-cdn.com/v4/letter/d/a8b319/32.png) [@dinesh](https://ask.bioexcel.eu/u/dinesh)\
**Post date:** [July 30, 2021, 5:55am UTC](https://ask.bioexcel.eu/t/haddock-z-score-vs-rmsd/3185/5 "2021-07-30T05:55:18Z")

</div>

sure will check them out. Thanks for your reply.