Structural Quality Assurance

1
Structural Quality Assurance

• Roman A. Laskowski

2
The importance of structures

• Individual structures increase understanding of
  specific function or of the mechanism of
  catalysation
• Multiple structures comparison between
  structures can reveal universal principles of
  structure-activity relationships

3
How to obtain a structure...

• By X-ray

• By NMR

4
Precision and accuracy
Precise Accurate Both!
5
To check for accuracy

• PDB files generally do not give standard
  uncertainties for atomic coordinates or
  structures in general
• Smaller molecules are often better defined - more
  precise
• Newer PDB entries are more likely to be close to
  the truth

6
X-ray precision parameters

• Resolution dependant on size
• R-factor difference between calculated model and
  raw data
• R-free a restricted R from a test-set, which is
  never refined
• Atomic B-value the disorder around the mean
  position, caused for example by alternate
  conformations

7
NMR precision parameters

• Root mean square-deviation (rmsd) is a factor
  related to inter-model agreement
• Number of restraints per residue increase
  likeliness of accuracy
• Dipolar coupling can be calculated for a given
  model, and compared to its data
• Cross-validated R equivalent to R-free

8
Yet another method
By looking at the stereo-chemistry and other
properties of the molecule, a model can be
compared to existing ones, providing a method of
validation independent of the original data
By only publishing the C? backbone of the
structure, many potential errors can simply be
evaded
9
Achtung an active site ought to be weird These
methods might weed it out, making it a boring
strand
10
Doesnt this just give perfect structures?
11
Four types of major errors

• Incorrect main- or sidechain configurations
• Frame shift errors
• The secondary-structure elements are incorrectly
  connected
• The path of the protein is totally wrong

12
Properly judging structures

• A Ramachandran-plot is a useful tool
• Torsion angles are clearly displayed
• Favoured areas are shown in colour
• Odd ones out can easily be picked

13
Where are the sidechains at?

• Ramachandran plots are limited to the backbone
• Sidechain torsion angles are displayed in ?1?2
  plots
• Favoured angles are indicated by the dashed lines

14
Other criteria

• Too much bad atom-atom contacts count against the
  model
• An excess of unsatisfied hydrogen-bond donors
  makes the model less plausible
• Cavities are thermodynamically unfavourable,
  hence improbable

15
Fortunately, were not alone

• Lots of software was written to help find errors
  in protein structures
• Most of these programs do not require specialist
  knowledge
• There are quite a lot of web-based servers where
  one can submit structures and retrieve quality
  reports

16
PROcheck

• PROcheck calculates a number of stereochemical
  parameters and gives graphical output files
• Ramachandran, ??, bond lengths and angles,
  planarity etc.

17
WHAT_check

• A subset of the WHAT_if package
• Checks a gargantuan number of properties
• Gives a very long and detailed output file
• The report contains less graphical data