Flips and flaws

29 Nov 2011

The PDB_REDO logo

The Protein Data Bank (PDB) contains approximately 70,000 protein structures and is an essential resource for anyone involved in structural biology. However, the quality of those structures is not always up to standards. Not surprising, considering the fact that some of these structures were calculated more than 30 years ago. To benefit from the progress in computational methods, the PDB_REDO project strives to enhance the quality of the structures by employing new methods to 'old' data. This has already resulted in numerous refinements or sometimes major revisions of published protein structures. Recently, new approaches have been incorporated in the PDB_REDO procedures, specifically targeting potential mistakes in the conformation of protein side chains (SideAide) and the protein backbone (Pepflip).

"In essence, with SideAide we can detect mistakes in side chain conformation by comparing simulated electron density maps to the electron density maps calculated from published data. When these show large differences, SideAide can rebuild the structure ", says Robbie Joosten (Netherlands Cancer Institute). SideAide starts out from the 'easy' residues, Joosten explains. "Residues like Valine and Proline have relatively little conformational freedom, so fitting these in is the easiest. Next we move to the more difficult residues, with Arginine being the final one. This residue is notorious for causing a lot of problems when building the structure."

With Pepflip, conformational mistakes or 'flips' in the protein backbone can be identified and corrected. Joosten: "Pepflip is a stepwise approach. First we go in with the 'big guns', like geometry analysis and coarse electron density map comparisons. When we find deviations, we analyse those peptide bonds in great detail. Although backbone flips are much less common than bad side chain rotamers, the consequences are more severe, especially when performing homology modelling."

Using the new methods on two-thirds of the structures in the PDB already resulted in an improvement of 80% of the structures. Even though they are really conservative in their approach, Joosten emphasises. "We only adjust a structure when we are certain that our calculations offer an improvement. You are dealing with other people's data and results, so you should be very careful."

All results and methods are available through PDB_REDO: http://www.cmbi.ru.nl/pdb_redo

RP Joosten, K Joosten, SX Cohen, G Vriend, A Perrakis
Automatic rebuilding and optimization of crystallographic structures in the PDB
Bioinformatics 2011, published online 27 October 2011

related publication (co-authored by RJ Joosten):
RJ Read et al,
A new generation of crystallographic validation tools for the Protein Data Bank
Structure 2011, 19(10):1395-1412