RMSD between re-docked complex and co-crystallized complex

Question

[ksraji251]

Hello,

I would like to find the rmsd between re-docked complex and co-crystallized complex for docking validation? What are the tools available for this purpose? I tried with pymol using align command, but I got the rmsd between two proteins not for complexes.

Answer 1

[jgreener]

You could try using MDAnalysis, see for example https://userguide.mdanalysis.org/stable/examples/analysis/alignment_and_rms/rmsd.html.

Answer 2

[Michael]

I remember doing this using the rmsd function in UCSF Chimera in this paper. To achieve that, I loaded the docking result and the original structure from PDB (that from which the ligand was deleted prior to docking). Then I deleted all protein residues for simplification. Then, RMSD was computed only taking C, N, and O into account, but to do this, all atoms needed to be aligned (as to appear as proper pairs) between the docking and the ligand. I think I did this manually all the time by superimposing the ligand structures and entering the atom identifiers. I feel there must be a better way but it worked for the few structures we had.

Edit: So, I found the helper script in R. To use that, one has to create a new PDB file for the purpose of aligning the atoms, don't overwrite your docking file. There, the compounds are superimposed such that all atoms are nearest neighbor of their respective image in the other ligand. The PDB should exactly contain two models, the docking ligand and the PDB ligand.

Then function align.nn.pdb gives the respective atom ids. The function name is a bit misleading because you would have done the alignment yourself already before.

Then, formatAs.chimeraAtomIds can convert the output into a string of correctly paired atom identifiers to use with RMSD.

If you think this is crude, I totally agree and please provide a better solution.

RMSD was sensitive for even slight rotation of the docking pose, so I think distance between centroids is also a good measure to provide.

Here is the Gist (pls, don't ask for details on how this even works :) )

Here is an example PDB file to try this on. One model contains ATOM's (the docked ligand) and the other (the original PDB model) HETATOM definitions, but I think that doesn't matter I think.