How to account for missing residues when docking small molecule ligands into a PDB protein crystal structure?

Question

[Rainer]

For a protein crystal structure from the PDB I would like rank small-molecules as putative binders to the protein's active site using a molecular docking software and subsequent binding affinity predictions for the best docking poses (I am currently using FlexX and HYDE for this purpose). The crystal structure has a good resolution, however some residues are missing as described in the comment section of the PDB-file - they are not close to the binding pocket, but I am worried that they could still affect induced-fit computations applied by the HYDE software.

I have created a new version of the crystal structure by adding the missing residues following the tutorial of the Modeller software for this purpose. However, this resulted in changes in the entire structure rather than only local changes.

My current strategy is to perform the docking simulations both with the original structure and with the output structure from Modeller, but I am not sure whether I can rely on the Modeller-derived structure at all. If you have alternative ideas on how to deal with this problem, I would greatly appreciate your suggestions.

Answer 1

[João Rodrigues]

Hi Rainer,

What you did sounds very logical and is appropriate. You are probably getting some side chains changed due to the molecular dynamics steps of Modeller, which tries to refine the structure. You can, if you want, restrict the refinement only to those residues you added: link.

Nevertheless, if your residues are VERY far away from the binding site, just ignore them. Even with induced fit, the refinement of the docking program is rarely good enough to take such distant effects correctly into account.