I am doing a docking study on viral protein. The protein structure which I downloaded from PDB is not reliable when I checked in the PDB validation tool. Can anyone suggest what approach I should take at this situation?
I used developmental version of the PDBe interface where they have implemented the validation tools
http://wwwdev.ebi.ac.uk/pdbe/entry/search/index
It depends on the type of problem with the existing structure. It's better to specify, in which way the deposited structure is not reliable? If it misses some residues - they can be modelled, if it has stereochemical problems - they can be resolved as well.
Look in detail at the validation report. The validation reports show the how various aspects of a structure deviate from the expected values, but these don't necessarily make a structure unreliable. Older structures have worse geometry than newer ones as refinement methods have improved, so they will show as poor in the validation reports, but it doesn't have to mean they are unreliable.
It sometimes only takes a few outliers to make the overall validation quality look poor, so look at the detailed validation report to see where the problematic residues are. For instance, if the Ramachandran outliers are throughout the protein, there may be more of a problem than if they cluster in flexible surface loops distant from the area you're interested in.
A structure might show poor fit to electron density, but if the residues which have a poor fit are on the surface and far from your region of interest, you don't care. If however the whole structure fits the density badly, that's an issue.
For docking,you're probably interested in surface residues and fit to density is critical. If for instance a long sidechain such as arginine is built protruding from the surface, but there is poor electron density for it, then it must be mobile, so docking with it the fixed position it has in the file would be wrong.
Remember also that the reports list deviations from the average or expected values, especially in geometry. A bond length/angle which is far from average might be like that for a very good reason, for instance if strained in an active site of an enzyme.
Unless the structure is absolutely horrendous, I would prefer one where you can look at the fit of the model to the experimental data (the electron density) than a homology model which has no experimental data.
Proteins are not rigid entities, so in docking you should ideally consider protein flexibility anyway. Any problems in the structure will most probably be accounted for if you use a docking tool that treats the binding pocket as flexible.
- Badges
- Science method