Van den Akker Lab-Structure Validation

Research

Home

DDQ (Difference Density Quality)

A new validation tool to assess the local and global correctness of macromolecular crystal structures

Focco van den Akker & Wim Hol

How to quantify great density like this?

Download color reprint (PDF file)

Download DDQ program (now CNS compatible)

Here is an example of a well refined region of a crystal structure (PDBid=1MYR) which is quite obvious from its hydrated difference map with the appearance of many strong water peaks (blue) and almost absence of negative shift-peaks density (red). The presence of correctly positioned strong water peaks and absence of shift peaks (as well as unmodeled parts) is actually a signature of a correct finished crystal structure and that signature can now be quantified for your own structure using DDQ to convince yourself and others of the soundness of the structure and to assess the completion of crystallographic refinement.

Aim of DDQ

DDQ is meant for providing a rigorous new validation method of crystal structures with the aim of obtaining the most accurate structure. This is especially desired for structures that are a starting point of rational drug-design in which a highly accurate model is needed to warrant the substantial investments.

DDQ is a novel crystallographic tool (using diffraction data) to assess:

the local correctness of macro-molecular crystal structures

the global correctness of macro-molecular crystal structures and will give you a ranking on how your structure compares to other PDB structure of similar resolution (ranking as follows: worst < bottom 25% < below average < above average < top 25% < best).
Please feel free to include this ranking in your crystallographic Table 1 dealing with refinement statistics.

whether crystallographic refinement has been completed satisfactorily (i.e. a. no shift peaks, b. many strong correctly positioned water peaks, c. absence of additional positive density due to unmodeled residues/ligands)

In addition, DDQ can also aid in refinement:

it sorts the residues with the worst shift-peaks scores first to direct rebuilding to where the largest disagreement is found between the model and the data thereby potentially allowing the refinement to converge faster

calculates map quality indicators for optimizing map parameters (such as resolution ranges, scale, SigmaA weights) to achieve the most informative difference map without visual inspection (potentially very helpful when you have only a partial model)

outputs thoroughly stereochemically-checked waters in .pdb format for use in refinement

monitors the presence of extraneous positive density features that could be due to missing parts of the model. These features are clustered and written into a .pdb file for easy inspection.

DDQ performance (download PDF file for more details)

DDQ has been tested on a large variety of different test structures and has been shown to be very sensitive to even small local errors in both low and high resolution structures

DDQ was able to pick up 10 errors in a randomly choosen deposited PDB structure with no false positives despite a 'very acceptable' R-factor of 18.9 % (2.4 Angstrom structure). The errors included a peptide flip, frame shift, incorrectly modeled side chains, and incorrect temperature factors.

The global DDQ scores are very sensitive to the average coordinate error and phase error

Omission of a small ligand is easily detected by DDQ; a feature which other structure validation programs have to date neglected to look for

Short summary of the principles of DDQ

DDQ is a new method for the automatic assessment of the local and global correctness of a macromolecular crystal structure using a difference Fourier electron density map, and the known preferred positions of electron density features due to the presence of waters. The method is based on the information found in |F_obs| - |F_calc| difference Fourier electron density maps in which water molecules are deliberately omitted from the structure factor calculation. Such difference Fourier electron density maps provide a wealth of information concerning the local and global accuracy of the crystal structure:

I) Hydrophilic, and to some extent also hydrophobic, atoms that are correctly positioned within the structure can make favorable interactions with positive water peaks in the difference electron density map. Analysis of such water peaks can be carried out using the detailed biophysical information known about the hydration of atoms in protein structures. The appearance of these water peaks not only is a (relatively) unbiased indicator for the quality of the atomic positions of the amino acid residues surrounding these waters but is also an indicator for the quality of the phases, and thus overall correctness, of the crystal structure.

II) Atoms that are either completely mis-positioned, slightly mis-positioned, or have an incorrect temperature factor or occupancy assigned will have negative and/or positive density peaks nearby, called 'shift peaks'. The degree to which such incorrect atoms can be detected in the difference electron density map depends on the phase error and thus on the quality of the rest of the structure.

III) Segments of the structure not included in the model but present in the crystal are likely to correspond to extra positive electron density at their atomic positions. Significant amounts of "unaccounted" positive density is an indicator of shortcomings of the model.

The distribution of the peaks in each of the three catagories listed above can be be used as a quantitative measure for completion of crystallographic refinement, i.e. very few peaks in category II & III, and many strong peaks in catagory I, something that the crystallographer always tries to aim for when refining a macromolecular crystal structure but the huge amount of 4 dimensional information found in a difference Fourier map (x,y,z, and peak height) can now be quantified automatically by DDQ to arrive at quality indicators for the structure.

How to obtain the DDQ software

The DDQ software and documentation can be obtained here, free of charge.

Click below to get a color copy of the full article as it appeared in Acta D:

DDQ reprint in PDF format (Acrobat Reader)(758 K)

DDQ reprint in level 2 Postscript format (7372 K)

Reference