standard deviations for cell constants
zbyszek at mix.swmed.edu
Tue Jun 11 13:00:05 EST 1996
Rob Hooft wrote:
> Zbyszek Otwinowski <zbyszek at mix.swmed.edu> writes:
> > Conclusion:
> > There is no sensible procedure for reporting
> > error of the unit cell. Fortunatelly it does not matter, as the error
> > is very small.
> However, after the refinement of a structure with inaccurate cell
> parameters, a systematic deviation of the bond lengths in the
> structure can be detected. The effect may be "marginal" in many cases,
> but in over 50% of all protein structures in the PDB it is detectable
> at the >99.99% confidence level! In some cases, the error is >2% in one
> of the axis lengths.
> In one particular case known to us, the crystallographer --worried by
> the warning message from our software-- found a problem in the dataset.
> Solving the problem resulted in a 1% drop in R.
The problem was probably due to a large error in reported
unit cell. Large error can be made in positional refinement
if one enters wrong detector parameters, e.g. distance.
Postrefinement (global refinement) is free from such problems.
The error in postrefinement is very small, however it is hard
to quantify how small it is. Standard error propagation in
postrefinement gives extremely small error estimates. Systematic
error result in larger actual error of postrefinement procedure,
however even systematic errors are unlikely to affect atomic
refinement. How big was the error in the unit cell that resulted
in 1% drom in R-factor? What was the resolution?
The unit cell errors are much larger in triclinic and monoclinic space groups.
In these space groups positional refinement at medium resolution can have
quite large error, 1% or larger.
Zbyszek Otwinowski | zbyszek at chop.swmed.edu
University of Texas | tel : (214)-648-5098
Southwestern Medical Center | fax : (214)-648-5095
5323 Harry Hines Boulevard Dallas, Texas 75235-9038
More information about the Xtal-log