Why HADDOCK Score doesn't have unit?

Dear all; as you know haddock score is: HADDOCK-score = 1.0EVDW+0.2EELEC+1.0*EDesolvation
and here the unit of Energy terms is kcal/mol and also 1.0, 0.2 and 1.0 are a weighted sum of intermolecular energies (vdw, elec, desolvation) but what haddock score does not have unit?

1 Like

The HADDOCK score is a combination of energy terms calculated using the OPLS force field (VDW, electrostatics), together with additional empirical terms (desolvation, buried surface area)) and restraint violation energies. Therefore, it can only be expressed in arbitrary units and cannot be used to predict binding affinities or compare different complexes. It should only be used to compare different solutions for a given complex.

1 Like

To add to that, in principle the final score after water refinement could have units, BUT we don’t want users to interpret it in terms of binding affinity, which is why we rather avoid putting units.

There is simply no correlation between docking scores in general (whatever software or method you use) and binding affinity. As such comparing the scores of various complexes does not make sense. For more information refer to our following publication:

1 Like

many thanks, so what is your suggestion to predict binding affinity to compare different complexes?

Try: http://milou.science.uu.nl/services/PRODIGY

See also the following relevant publications:

And there is next week a BioExcel webinar about this topic!

This has been very helpful for me to. I was wondering however if you have an suggestions for predicting binding affinities in protein-DNA complexes ?

No real suggestions here I am afraid…
Would have to search literature for it, which you can also do :slight_smile:
But it is on our radar.

Do these things apply to protein-ligand docking too??

The simple answer is yes!

See for example our recent D3R paper, which does include ranking of various ligands for one receptor.
The HADDOCK score does not do good job at it… But we developed a simple contact-based predictor instead.

Z. Kurkcuoglu, P.I. Koukos, N. Citro, M.E. Trellet, J.P.G.L.M. Rodrigues, I.S. Moreira, J. Roel-Touris, A.S.J. Melquiond, C. Geng, J. Schaarschmidt, L.C. Xue, A. Vangone and A.M.J.J. Bonvin.
Performance of HADDOCK and a simple contact-based protein-ligand binding affinity predictor in the D3R Grand Challenge 2.
J. Comp. Aid. Mol. Des. 32, 175-185 (2018).

And we should soon have results of the Grand Challenge 3…