Page 197 - Macromolecular Crystallography

186 MACROMOLECULAR CRYS TALLOGRAPHY

information from the database. These data are structure factors by Truncate (also from CCP4).
then provided to the program, marccd, which Processing for each dataset is assigned to one of
operates the sample changing robot and the detec- nine servers located at the beamline. Reduced
tor system. This data file also includes proto- data are sent over the T1 line to SGX San Diego,
col commands for turning the liquid nitrogen where further processing occurs on a Linux cluster
wash on and off, archiving of data, and initia- containing 296 processors.
tion of the crystal quality evaluation system. The
entire carousel load is then screened automati- Most of the data collected at SGX-CAT represent
cally. For each crystal, the nylon mounting loop cocrystal of ligands with proteins of known struc-
is automatically centred, after which four screen- ture. Cocrystal structures are determined via molec-
ing diffraction images are recorded at orientations ular replacement. EPMR (Kissinger et al., 1999),
◦
◦
◦
◦
with φ = 0 ,45 ,90 , and 135 . Once these MOLREP (Vagin and Teplyakov, 1997), or Phaser
images have been acquired, the crystal scoring (Read, 2001) are used to find the best molecular
system then analyses the images, determining replacement solution. Refinement of the structure
diffraction quality and establishing whether or and docking of bound ligands is accomplished using
not that crystal diffracts to the desired resolu- RefMac5 (Murshudov et al., 1997) or CNX (Brünger,
tion limit. All of the parameters generated from 1992).
this evaluation process are stored permanently in Alternative scripts are available for de novo struc-
the database for future use. Once quality scoring ture determination. Initial processing usually uses
has been completed for all of crystals within a the Laue class determined during crystal evalua-
given set of replicate samples, the best crystal is tion. The phasing method depends on the type of
marked automatically for collection on the status experiment used to generate the data. In the case
web page. of a MAD/SAD dataset, ShakeNBake (Hauptman,
Collection: Collection of datasets is executed in much 1997) or SHELXD (Schneider and Sheldrick, 2002) is
the same way as screening. In order to maximize used to determine the locations of anomalous scat-
the use of beam time while minimizing manual terers. MLPHARE (Otwinowski, 1991) (heavy atom
effort, new carousels of crystals are created that parameter refinement and phasing) and SOLOMON
contain only crystals deemed suitable for data col- (Abrahams and Leslie, 1996) or DM (Cowtan,
lection. At this stage, the crystals are identified 1994) (phase improvement using solvent flatten-
through the 2D barcodes located on the base of ing/flipping) are used to generate and improve the
the pin. Sample mounting, liquid nitrogen wash- phases based on the anomalous signal present in the
ing, and automated centring are performed dur- data. Once the phases have been refined to conver-
ing data collection as during screening. Optimal gence, ARP/wARP (Lamzin and Wilson, 1993) is
sample-to-detector distance for each sample is cal- used to build the model, followed by visual inspec-
culatedfromtheestimatedresolutionlimitandthe tion and rebuilding with Xfit (McRee, 1999) or COOT
X-ray wavelength. Protocol files for data collec- (Emsley and Cowtan, 2004), and further refinement
tion also include the oscillation size and number with RefMac5. When only relatively low resolution
of diffraction images specified in the database, data (2.5–3.5 Å) are available, MAID (Levitt, 2001)
both of which are determined automatically from is used instead of ARP/wARP to build the initial
knowledge of the sample Laue point group. atomic model.
Data processing: Immediately upon completion of the
acquisition of a diffraction dataset, data process- Validation: Once a structure has been determined,
ing is launched as part of the process for changing it is validated using a custom structure valida-
to the next sample. At SGX-CAT, data are indexed tion system (Badger and Hendle, 2002) to detect
and integrated using either Mosflm or d*TREK. local errors. The system is based on PROCHECK
Scaling of the integrated data is performed using (Morrisetal., 1992), WHATCHECK(Vriend, 1990),
Scala from CCP4 (Collaborative Computational SFCHECK (Vaguine et al., 1999), and PHISTATS
Project, 1994). Scaled intensities are converted to and OVERLAPMAP (from CCP4).