Dear all,
The newly-deposited structure 3ftt from the CSGID structural genomics
group has fields in its sf.cif file called F_meas_au and F_meas_sigma_au,
but which clearly contain F_calc and ph_calc - at least, the F_meas_sigma_au
field is uniformly distributed on [0,360] and the F_meas_au field has a >90%
correlation with the calculated structure factors from the deposited PDB
file in all resolution bins.
Thomas Womack's routine re-refinement of all structures deposited each
week picked this up through a final Rfree value of 47% and a difference map
with points at both +35 sigma and -35 sigma. His manual inspection of the
difference map indicated negative density at every oxygen and nitrogen atom,
and a more in-depth look at the structure factors picked up the root cause.
The link between that cause and its effect on the final Rfree was explained
by the fact that the erroneous sigma values resulted in very small F/sig(F)
values at high resolution, hence in a very high value of the refined Luzzati
B factor produced by BUSTER-TNT, and an anticipated decorrelation between
expected and observed amplitudes. The attached reciprocal-space correlation
coefficient plot showed this very clearly.
This instance leaves room for some concern about the verification of
deposited structure factors: what happened here is consistent with the
hypothesis that the standard-deviation field is ignored entirely, or at
least that no plausibility check is carried out on the values deposited in
it.
With best wishes (and thanks to Thomas Womack),
Gerard.
--
===============================================================
* *
* Gerard Bricogne g...@globalphasing.com *
* *
* Global Phasing Ltd. *
* Sheraton House, Castle Park Tel: +44-(0)1223-353033 *
* Cambridge CB3 0AX, UK Fax: +44-(0)1223-366889 *
* *
===============================================================
