Protein–Glycan Docking

HADDOCK
1

Hello,
I’m a PhD student recently getting into protein–glycan docking. I’ve started using the HADDOCK 2.4 webserver and have been running protein–glycan docking with oligosaccharides generated using GLYCAM. So far, the runs have been working fine.

My main question is how to interpret some of the results. I already know several proteins that bind to this glycan, and I’m now assessing other potential targets.

I’d like to know whether comparing HADDOCK scores is meaningful. For example, if proteins not previously known to bind this glycan yield similar scores to known binders, could that suggest potential interaction?

Also, is there any way to assess the affinity of glycan–protein binding based on docking results?

Thank you for your support.
Best,
Amirali

2

Hi Amirali,

comparing HADDOCK scores is meaningful only when the docked molecules are similar to each other (the more of the system remains constant the more reliably you can compare scores). so it depends on the potential targets.

But in general you should not use the HADDOCK score to predict interaction. And the score is not equal to binding affinity. For that you can try to apply other predictors on the best scoring models you obtain with HADDOCK!

PS: the forum is searchable and there’s a lot of information about possible comparisons between docking scores :wink:

Hope it helps,
Marco

3

And to add to Marco’s answer, I am not aware of a binding affinity predictor for protein-glycan complexes.

Reliable data and their amount is most likely the limiting factor here to train any model.

4

Thank you for your prompt response.

I also haven’t encountered any binding affinity predictors specifically for protein–glycan complexes, and I understand such docking studies remain limited.

To confirm and reiterate what I gathered from your previous responses and the forum:

  • HADDOCK scores are best used for comparing clusters within a single docking run.
  • HADDOCK scores can be compared when docking different ligands to the same protein.
  • The similarity of docked molecules determines whether such comparisons are valid.

Given these considerations, I have some additional questions regarding our study design:

We have a set of candidate proteins selected based on their isoelectric points (pI), implying possible electrostatic interactions (positive/negative charges) with our glycan. To move beyond predictions based solely on charge, we’re interested in performing docking studies to identify promising candidates for in vitro validation.

Specifically:

  • Regarding the necessary similarity for meaningful HADDOCK score comparisons: we are docking the same glycan (constant ligand) to several proteins belonging to two family. When you mention similarity, do you refer primarily to structural similarity, sequence identity, or similarity in active-site residues?
  • Since direct HADDOCK score comparisons between different proteins isn’t advisable, would you recommend using alternative metrics (e.g.such as Van der Waals energy, electrostatic energy) for comparisons across different proteins?
  • Would docking a known non-binding control protein help establish a meaningful baseline for identifying negative interactions with HADDOCK?

Thank you again for your guidance.

Amirali

5

We have a set of candidate proteins selected based on their isoelectric points (pI), implying possible electrostatic interactions (positive/negative charges) with our glycan. To move beyond predictions based solely on charge, we’re interested in performing docking studies to identify promising candidates for in vitro validation.

Specifically:

  • Regarding the necessary similarity for meaningful HADDOCK score comparisons: we are docking the same glycan (constant ligand) to several proteins belonging to two family. When you mention similarity, do you refer primarily to structural similarity, sequence identity, or similarity in active-site residues?

Actually all three :slight_smile: And mainly for the binding site as this will be the region contributing mainly to the HADDOCK score.

  • Since direct HADDOCK score comparisons between different proteins isn’t advisable, would you recommend using alternative metrics (e.g.such as Van der Waals energy, electrostatic energy) for comparisons across different proteins?

The HADDOCK score is a combination of those energies. So won’t make much difference.

  • Would docking a known non-binding control protein help establish a meaningful baseline for identifying negative interactions with HADDOCK?

Yes for sure - this is a research project in its own. You should then also dock known binders. I.e. a benchmark of positive and negative cases.