Page 131 - Macromolecular Crystallography

120 MACROMOLECULAR CRYS TALLOGRAPHY

of the f curve from its theoretically calculated posi- 8.3.3 Data measurement and processing
tion (Fanchon and Hendrickson, 1990). It is also
Dataprecisionandaccuracyaregenerallyimportant,

clear, from Fig. 8.1, that the f curve varies very
but acquire additional importance due to the small
rapidly, as a function of wavelength, near the inflec-
size of MAD signals. Although some multiwave-
tion point. Since the magnitude of the measured
length data have been measured on laboratory
dispersive signal is dependent on the difference in f
sources (Hendrickson et al., 1986), synchrotron beam
values between the inflection point and the ‘remote’
lines are the most practical sources for such mea-
wavelength chosen, prior accurate determination
surements. Signals at K absorption edges tend to

and representation of the f curve is important. The
be numerically smaller than those at the L and M

f and f values are generally determined from flu-
edges (Fig. 8.1) and somewhat larger errors can be
orescence measurements that are then transformed
tolerated in the latter cases. When data measure-

into f and f values through use of programs such
ments were usually made with crystals mounted
as CHOOCH (Evans and Pettifer, 2001) that use
within capillaries, it was customary to enhance
standard transformations (Cromer and Lieberman,
precision in measured Bijvoet mates by orienting
1970) for the purpose. Wavelengths at which to
crystals such that reflections related through Friedel
measure data can then be directly read off from f
symmetry were measures on the same frames, at

and f curves plotted as a function of wavelength.
nearly the same time (Murthy et al., 1988, 1999).
For measurement of Bijvoet signals the choice of
However, with flash-frozen crystals, orientation
wavelength is less critical, at least if one ascertains
along real or reciprocal crystal axes is not practi-
that the monochromator is positioned on the lower
cable. Thus in generally for data measurement on
wavelength side of the absorption curve, as the
flash-frozen crystals the inverse-beam geometry is
absorption continues to be high at longer distances
preferred (Hendrickson, 1991). The time required
from the peak value (Fig. 8.1). Acquisition of flu-
for movements of the goniometer motor to mea-
orescence scans, derivation of the absorption and
sure data in the inverse-beam geometry would
dispersion curves, and suggestion of wavelengths
worsen potential radiation damage to the crys-
for data measurement are completely automated at
tal. Although measurement of data on flash-cooled
most synchrotron beam lines. There are also web-
crystals has reduced the severity of radiation dam-
sites, such as that at the University of Washington
age it is still a significant problem, even at –100 C
◦
(Merritt, 1998), which permit theoretical estimates
(Garman, 2003). In most cases this conundrum is

of f and f values for potential anomalous scatter-

◦
addressed by measuring data in blocks of 30 –45 ◦
ers to be obtained, as well providing other useful
before swinging the goniometer to the correspond-
information for experimental design. In addition,
ing inverse orientation, and alternating until all
other prescriptions for making optimal wavelength
the data needed are measured. If needed, even
choices are available (Narayan and Ramaseshan,
smaller blocks can be used for crystals that show
1981; Burla et al., 2004).
little decay. A review by Gonzalez (2003) has greater
As a practical example, the theoretical spectrum
detail on data measurement principles and practices,
and the wavelength choices made for measurement
including measurement of MAD data. Processing
of MAD data for vaccinia complement-control pro-
of data at each wavelength is done as it is for sin-
tein (VCP) are shown in Fig. 8.2. There were two
gle wavelength experiments, with the proviso that
crystal forms of the protein, in space groups P2 1 2 1 2 1
the Bijvoet differences are not merged, at least at the
and C222. One heavy atom derivative, Eu +++ and
peak wavelength.
Au +++ , respectively, for each of them was obtained
The most popular processing programs in
and MAD data were collected on each of them
current use are probably DENZO/SCALEPACK
on the X4A beam line at NSLS. The structure was
(Otwinowski and Minor, 1997) and MOSFLM
eventually determined, through MAD phasing for
(Project, 1994). Careful scaling is particularly
the P2 1 2 1 2 1 crystal form and the structure in C222
important to preserve the relatively weak anomal-
determined using molecular replacement (Murthy
ous signals, and minimize the consequences of
et al., 2001).