Gerard Kleywegt does not take kindly to dodgy PDB files. I've been aware of his work in a general sense for a while, but after reading Practical Fragment's review of "Limitations and lessons in the use of X-ray structural information in drug design" by Andrew Davies, Stephen St-Gallay and Gerard Kleywegt I decided to read some of his papers. If you are looking for a broad overview (with excellent examples) of the issues to be considered if using PDB files, this paper is a good place to start.
The PDB is a wonderful resource, but it should be remembered that crystallography provides the electron density (at a specific resolution) not the structure, and everything from that point on is modelling. In other words, mistakes can be made due either due to ambiguities or incorrect assumptions. In most (?) PDB files, the electron density can be reconstructed to some extent from structure factors, and compared to the model to find regions where there is a lack of density, and hence a lack of evidence. In particular, the ligand can get very little love from the crystallographer (see references 43-48 in the paper).
I see that Kleywegt, after 15 years at Uppsala, has recently joined the Macromolecular Structure Database Group at the EBI as head of the PDBe, the European end of the PDB. Does this mean that dodgy PDB files will be a thing of the past?
Image credit: Ethan Hein