2QVO.xds

XDS data reduction

dataset 2

This is a pared-down XDS.INP (obtained by egrep -v '^ *!' XDS.INP) based upon XDS-MARCDD.INP from the XDS distribution site - it has only those lines that are not commented out (to arrive here, one takes the steps outlined in Tutorial(First_Steps)):

DETECTOR=CCDCHESS        MINIMUM_VALID_PIXEL_VALUE=1     OVERLOAD=65000
DIRECTION_OF_DETECTOR_X-AXIS= 1.0 0.0 0.0
DIRECTION_OF_DETECTOR_Y-AXIS= 0.0 1.0 0.0
TRUSTED_REGION=0.0 0.99 !Relative radii limiting trusted detector region
MAXIMUM_NUMBER_OF_PROCESSORS=8!<25;ignored by single cpu version of xds
JOB=  XYCORR INIT COLSPOT IDXREF DEFPIX  INTEGRATE CORRECT
ORGX=2000 ORGY=2048  !Detector origin (pixels)! numbers are rough estimates w/ adxv
DETECTOR_DISTANCE= 125.0   !(mm)
ROTATION_AXIS= 1.0 0.0 0.0
OSCILLATION_RANGE=1.0            !degrees (>0)
X-RAY_WAVELENGTH=1.9         !Angstroem
INCIDENT_BEAM_DIRECTION=0.0 0.0 1.0
FRACTION_OF_POLARIZATION=0.95 !default=0.5 for unpolarized beam
POLARIZATION_PLANE_NORMAL= 0.0 1.0 0.0
SPACE_GROUP_NUMBER=0   !0 for unknown crystals; cell constants are ignored.
FRIEDEL'S_LAW=FALSE !Default is TRUE.
NAME_TEMPLATE_OF_DATA_FRAMES=../../g/040707-8_2_2_1.????  ! TIFF
DATA_RANGE=1 360       !Numbers of first and last data image collected
BACKGROUND_RANGE=1 5  !Numbers of first and last data image for background
SPOT_RANGE=1 180       !First and last data image number for finding spots
REFINE(IDXREF)=BEAM AXIS ORIENTATION CELL DISTANCE
REFINE(INTEGRATE)=DISTANCE BEAM ORIENTATION CELL !AXIS
REFINE(CORRECT)=DISTANCE BEAM ORIENTATION CELL AXIS
VALUE_RANGE_FOR_TRUSTED_DETECTOR_PIXELS= 6000 30000 !Used by DEFPIX for excluding shaded parts of the detector.
INCLUDE_RESOLUTION_RANGE=50.0 0 !Angstroem; used by DEFPIX,INTEGRATE,CORRECT
MINIMUM_ZETA=0.1 !Defines width of 'blind region' (XPLAN,INTEGRATE,CORRECT)
WFAC1=1.5  !This controls the number of rejected MISFITS in CORRECT; a larger value leads to fewer rejections.
STRONG_PIXEL=6.0                              !used by: COLSPOT

Using the above as XDS.INP, we run xds_par for the first time. It will stop after the IDXREF step with the usual error message

!!! ERROR !!! INSUFFICIENT PERCENTAGE (< 70%) OF INDEXED REFLECTIONS
AUTOMATIC DATA PROCESSING STOPPED. AS THE CRITERIA FOR A GOOD
SOLUTION ARE RATHER STRICT, YOU MAY CHOOSE TO CONTINUE DATA
PROCESSING AFTER CHANGING THE "JOB="-CARD IN "XDS.INP" TO
"JOB= DEFPIX INTEGRATE CORRECT".
IF THE BEST SOLUTION IS REALLY NONSENSE YOU SHOULD FIRST HAVE
A LOOK AT THE ASCII-FILE "SPOT.XDS". THIS FILE CONTAINS THE
INITIAL SPOT LIST SORTED IN DECREASING SPOT INTENSITY. SPOTS
NEAR THE END OF THE FILE MAY BE ARTEFACTS AND SHOULD BE ERASED.
ALTERNATIVELY YOU MAY TRY DIFFERENT VALUES FOR "INDEX_ORIGIN"
AS SUGGESTED IN THE ABOVE LISTING.
IF THE CRYSTAL HAS SLIPPED AT THE BEGINNING OF DATA COLLECTION
YOU MAY CHOOSE TO SKIP SOME OF THE FIRST FRAMES BY CHANGING
THE "DATA_RANGE=" IN FILE "XDS.INP" AND START ALL OVER AGAIN.

We choose to continue nevertheless and modify XDS.INP to have

JOB=  DEFPIX  INTEGRATE CORRECT

Again we run xds_par. This runs to completion. The automatic spacegroup determination comes up with

SPACE_GROUP_NUMBER=   75
UNIT_CELL_CONSTANTS=    53.10    53.10    40.90  90.000  90.000  90.000

Now we copy these two lines to XDS.INP, replacing the old line SPACE_GROUP_NUMBER=0 . Then we modify the spacegroup number to 77 because we know that the true spacegroup is P4_2. Also, we modify the JOB line once again:

JOB= CORRECT

and run xds_par for the last time.

The resulting output files are XYCORR.LP, INIT.LP, COLSPOT.LP, IDXREF.LP, DEFPIX.LP, INTEGRATE.LP and CORRECT.LP. Data files are XPARM.XDS (from IDXREF), and the XDS_ASCII.HKL file all of which can be downloaded from here (right-click with the mouse, and then save the file to your disk).

dataset 1

This works exactly the same way as dataset 2, except that we have to replace ../../g/040707-8_2_2_1.???? by f/040707-8_2_2_1.???? where f points to the directory with the frames. All .LP files, XPARM.XDS and XDS_ASCII.HKL are here (right-click).

SHELXC/D/E structure solution

This is done in a subdirectory of the XDS data reduction directory. Here, we generate XDSCONV.INP (I used MERGE=TRUE, sometimes the results are better that way) and run xdsconv and SHELXC:

 #!/bin/csh -f

cat > XDSCONV.INP <<end
INPUT_FILE=../XDS_ASCII.HKL
OUTPUT_FILE=temp.hkl SHELX
MERGE=TRUE
FRIEDEL'S_LAW=FALSE
end

xdsconv 

shelxc j <<end
SAD   temp.hkl
CELL 53.10 53.10 40.90 90 90 90
SPAG P42
MAXM 2
end

This writes j.hkl, j_fa.hkl and j_fa.ins. However, we overwrite j_fa.ins now:

cat > j_fa.ins <<end
TITL j_fa.ins SAD in P42
CELL  0.98000   53.10   53.10   40.90   90.00   90.00   90.00
LATT  -1
SYMM -Y, X, 1/2+Z
SYMM -X, -Y, Z
SYMM Y, -X, 1/2+Z
SFAC S
UNIT   128
SHEL 999 3.0
FIND 3
NTRY 100
MIND -1.0 2.2
ESEL 1.3
TEST 0 99
SEED 1
PATS
HKLF 3
END
end

shelxd j_fa

This gives best CC All/Weak of 35.61 / 26.03 for dataset 2, and best CC All/Weak of 36.74 / 21.55 for dataset 1.

Next we run G. Sheldrick's beta-Version of SHELXE Version 2009/4:

shelxe.beta j j_fa -a6 -q -h -s0.55 -m20 -b 

Some important lines in the output: for dataset 2, I get

   79 residues left after pruning, divided into chains as follows:
A:  20   B:  22   C:  37

CC for partial structure against native data =  50.42 %
...
 <wt> = 0.300, Contrast = 0.731, Connect. = 0.817 for dens.mod. cycle 20
...
Estimated mean FOM = 0.659   Pseudo-free CC = 68.71 %

for dataset 1, I get

   80 residues left after pruning, divided into chains as follows:
A:  23   B:  57

CC for partial structure against native data =  45.79 %
...
<wt> = 0.300, Contrast = 0.711, Connect. = 0.812 for dens.mod. cycle 20
...
Estimated mean FOM = 0.611   Pseudo-free CC = 63.70 %

clearly indicating that the structure can be solved with each of the two datasets individually.

For completeness, we run the inverse hand:

shelxe.beta j j_fa -a6 -q -h -s0.55 -m20 -b -i

but of course this gives much worse statistics.

Optimization of data reduction

The only safe way to optimize the data reduction is to look at external quality indicators. Internal R-factors, and even the correlation coefficient of the anomalous signal are of comparatively little value. A readily available external quality indicator is CC All/CC Weak as obtained by SHELXD.

WFAC1 was already discussed above. Another candidate for optimization is MAXIMUM_ERROR_OF_SPOT_POSITION. By default this is 3.0 . In the case of these data, this default appears to be too small, because the STANDARD DEVIATION OF SPOT POSITION (PIXELS) (as reported by IDXREF, INTEGRATE and CORRECT after refinement) is quite high (1.5 and more). This prevents XDS from using all the reflections for geometry refinement.

I found that MAXIMUM_ERROR_OF_SPOT_POSITION=6.0 significantly improved the internal statistics (mostly the r-factors, but not so much the correlation coefficient of the anom signal), and improved CC All/CC Weak indicators (to more than 40). SHELXE then produces significantly better and more complete models. Try for yourself!

One thing I noticed that if I specify the known spacegroup in IDXREF then the results are worse than if the integration is performed in P1. Likewise, optimization did not work: recycling of GXPARM.XDS to use as XPARM.XDS, and thus imposing the lattice symmetry in the geometry refinement in INTEGRATE. These findings my correspond to the fact that in P1 the angles do not refine to 90.0xx or 89.9xx degrees. In other words, the metric symmetry is not as well fulfilled as it should be. This results in fairly large deviations from the ideal P42 positions; the refinement of cell parameters in P1 partly compensates for this. I have however no idea why this deviation from metric symmetry could occur.

Optimization of structure solution

There are some parameters in the SHELXC/D/E approach above that could be optimized as well: first of all, MERGE=TRUE in XDSCONV.INP turned later out to be the wrong choice (using the default MERGE=FALSE turns out to give a model with 85 consecutive residues for dataset 1). Then of course, the resolution limit for SHELXD could be varied, and the solvent content for SHELXE. For SHELXE in particular, many things could be tried.

Limits

With dataset 2, I tried to use 270 frames but could not solve the structure using the above SHELXC/D/E approach (not even with MAXIMUM_ERROR_OF_SPOT_POSITION=6.0). With 315 frames, it was possible.