Hi,
Could anybody enlighten me about the current best practice
for calculating the e.s.u for atomic positions of individual atoms
based on B-factor, R-free as well as completeness of data and model.
I have been digging through the literature to find a good estimator
of the estimated standard uncertainties of atomic positions.
Most of these discussions lead to the so-called DPI developed
by Cruickshank.
(I think this stands for diffraction-data precision indicator).
Cruickshank gives a formula for it in his article in Acta Cryst.
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=10089455
But in this article he indicates that the B-factor dependence of his
formula
varies from protein to protein. He then gives parameters
for second-order polynomials to describe this B-factor dependence
for a few proteins.
Then there is a DPI defined in a CCP4 newsletter article
http://www.ccp4.ac.uk/newsletters/newsletter33/murshudov.html
equation 8 gives a B-factor dependent formula, which seems
to be general, but due to strong pixelation it is hard to make out
the details of that formula.
Finally there is a paper by Stroud and Fauman (Protein Science 1995)
that deals with the same issue, but with a rather different sort of
approach.
Finally different refinement programs give rather different R-factors
and
B-factors that can change the estimated s.u rather dramatically.
Any advice or pointers to further articles are wellcome.
Thanks,
Ulrich
