more important than the absolute values of CCall/CCweak is the
distribution of these statistics among the solution-trials of one SHELXD
computation job (i.e. based on a particular data set).
In the HKL2MAP scatter-plot, there should be at least one, or a set of,
solutions that are clearly separate in both CCall and CCweak from the
bulk of trials - representing a bimodal distribution. Importantly, the
best solution is promising if it is maximal in *both* CCall and CCweak.
On the contrary, to my experience, two "best" solutions that are apart
from the rest, but one maximal in CCall and the other in CCweak, point
at the presence of not-completely-wrong, but incomplete substructures
(in terms of correct sites).
Absolute "threshold values" for CC are hard to define, since these
statistics depend on various factors, among these importantly the data
resolution. But with all caution and no guarantee, I would start to call
SAD solution trials "promising", if their CFOM = CCall + CCweak exceeds 30%.
I am trying to solve a structure using Fe SAD collected with mini-kappa from several
non-isomorphic crystals (following on from the previous
Xtriage suggests weak anomalous signal to 4 A.
For the SHELX pipeline using hkl2map, are there any metrics/numbers that I can
use to judge the quality of each dataset? For example, what should be the value
of CCall/CCweak to decide between strong/promising and useless cases?
Is there any difference between using CRANK2 and hkl2map, considering that the
underlying software are the same?
Will the wavy Rmerge affect the 'solvability'?
Dr. rer. nat. Fabio Dall'Antonia
European Molecular Biology Laboratory c/o DESY
Notkestraße 85, Bldg. 25a
phone: +49 (0)40 89902-178
fax: +49 (0)40 89902-149