From XDSwiki
Revision as of 10:30, 26 November 2018 by Kay (talk | contribs) (Anomalous Fcalc as reference (under development))

XDSCC12 is a program for generating delta-CC1/2 and delta-CC1/2ano values for XDS_ASCII.HKL (written by XDS), or for XSCALE.HKL containing several files of type XDS_ASCII.HKL after scaling in XSCALE (with MERGE=FALSE).

It implements the method described in Assmann, Brehm and Diederichs (2016) Identification of rogue datasets in serial crystallography. J. Appl. Cryst. 49, 1021 [1], and it does this not only for the individual datasets in XSCALE.HKL, but also for individual frames, or groups of frames, of a single dataset collected with the rotation method and processed by XDS.

The program can be downloaded for Linux 64bit or Mac.

Usage (this text can be obtained with xdscc12 -h):

xdscc12 KD 2018-9-6. Academic use only; no redistribution. -h option shows options.
Please cite Assmann, G., Brehm, W., Diederichs, K. (2016) J.Appl.Cryst. 49, 1021-1028
usage: xdscc12 [-dmin <lowres>] [-dmax <highres>] [-nbin <nbin>] [-mode <1 or 2>] [-<abcdeftwz>] [-r <ref>] FILE_NAME
dmax (default 999A), dmin (default 1A) and nbin (default 10) have the usual meanings.
mode can be 1 (equal volumes of resolution shells) or 2 (increasing volumes; default).
   -t: total oscillation (degree) to batch fine-sliced frames into
   -r: next parameter: ASCII reference file with lines: h,k,l,Fcalc
 FILE_NAME can be XDS or XSCALE reflection file
 other options can be combined (e.g. -def), and switch the following off:
   -a: individual isomorphous summary values
   -b: individual (Fisher-transformed) delta-CC1/2 values
   -c: individual delta-CC1/2 reflection numbers
   -d: individual anomalous summary values
   -e: individual (Fisher-transformed) delta-CC1/2ano values
   -f: individual delta-CC1/2ano reflection numbers
   -w: weighting of intensities with their sigmas
   -z: no Fisher transformation of delta-CC1/2 values

The program output is terse but supposed to be self-explanatory. The isomorphous delta-CC1/2 of a batch of frames (width chosen with the -t option) relative to all data is most easily visualized via XDSGUI (Statistics tab); the anomalous delta-CC1/2 may be plotted with e.g. gnuplot after grepping the relevant lines from the output.

For multiple datasets, the output lines show the contribution of each dataset toward the total CC1/2. Negative numbers indicate a worsening of the overall signal.

Statistics are given (in resolution shells) for the isomorphous and the anomalous signal.

Important: to identify outliers in XSCALEd data, you should use the -w option. Otherwise, a and b are adjusted such that the sigmas are very high, which reduces the delta-CC1/2 signal.

A complete description of how to process serial crystallography data with XDS/XSCALE is given in SSX. A program that implements the method of Brehm and Diederichs (2014) and theory of Diederichs (2017) is xscale_isocluster.

Preparing a reference data set

If the refinement was done with phenix.refine, one could use

mtz2various hklin 2bn3_refine_001.mtz hklout temp.hkl <<eof
LABIN FC=F-model PHIC=PHIF-model

- the column corresponding to PHIC will not be used by xdscc12. Alternatively,

read mymodel_001.mtz
write temp.hkl format(3i5,f10.3) col F-model

Fcalc with anomalous signal as reference

A simple way to obtain Fcalc(+) and Fcalc(-) is to run phenix.refine with options (in case of S as anomalous scatterer)

refinement.refine.anomalous_scatterers.group.selection="element S"

and then

sftools <<eof
read mymodel_001.mtz
write anom-reference.hkl format(3i5,2f10.3) col "F-model(+)" "F-model(-)"

in which case sftools outputs only the acentric reflections - only those have anomalous differences. XDSCC12 then has to be run with the -s -r anom-reference.hkl option. This allows a direct comparison of CC Iobs-Icalc, with Icalc either as (Fcalc(+) + Fcalc(-))/2 (these are the output lines starting with r) or with Iobs(+) being compared with Fcalc(+)2, and Iobs(-) being compared with Fcalc(-)2 (output lines starting with s). Preliminary insight: the r and s lines do not differ significantly.