Page 115 - Macromolecular Crystallography

104 MACROMOLECULAR CRYS TALLOGRAPHY

by conjugate-gradient minimization of the best solu- by Lewis and colleagues (Chang and Lewis, 1997)
tions (Jamrog et al., 2003). and the other resulted in the popular program EPMR
• Systematic rotation followed by translation (Kissinger et al., 1999, 2001). The EPMR program is
searches, as generated by a script (Sheriff et al., 1999) now widely used in the crystallographic community
(see also Protocol 7.6). and is especially simple to use (see Protocol 7.4). This
makes it a very attractive candidate for use in a suite
of programs integrated in a web site that goes all the
Alternatively, one might want to search the N-6D
way from model generation to the refinement of the
space (alpha, beta, gamma, tx, ty, tz for each one
final model (Rupp et al., 2002).
of the N molecules of the asymmetric unit) using
All these methods exploit the fact that there is no
stochastic or Monte Carlo methods, as exhaustive
need to recalculate the structure factors of the model
searches are out of question. In this case, as all
each time it is rotated or translated; it is sufficient to
molecules are searched simultaneously, the problem
be able to sample the structure factors at the rotated
of the low signal is less severe than with traditional
Miller indices, with or without a phase shift com-
MR methods. However, the process is quite cpu
ing from the translation, and this can be done quite
intensive. This has been implemented with success
effectively by interpolation in reciprocal space (see
by Glykos and Kokkinidis (Glykos and Kokkinidis,
Protocol 7.4).
2000, 2001, 2003), who later included a simulated
annealing protocol to increase the radius of con-
vergence of the method (Queen of Spades or Qs 7.8 How to choose the best model
method).
Other methods have used genetic algorithms to Even the best possible MR package will fail if
search the 6D space: one of them was originated the model is not good. Hence, a good deal of

Protocol 7.4 Stochastic search methods
Use of EPMR (Kissinger et al., 1999, 2001) TARGET CORR-1
epmr -m 2 -h 4. -l 12. -n 50 example.cell example.pdb CYCLES 10
example.hkl >example.log & STEPS 1000000
where example.cell contains the cell parameters and the STARTING_T 0.0150
space group number: FINAL_T 0.0050
INFO 1000
70.315 80.852 90.31 90. 90. 90. 19
NOISE_ADDED 0.10
and the model and X-Ray data are in example.pdb and RESOLUTION 12.0 4.0
example.hkl, respectively. AMPLIT_CUTOFF 5.0
The options -l and -h define the low and high resolutions SIGMA_CUTOFF 0.0
limits of the data, respectively, while the -m option defines RANDOM_SELECT 1.0
the number of molecules to be searched. FREE 0.10
The number of different starts is controlled by the -n MODEL example.pdb
option. DATA example.hkl
It is clear from this input lines that EPMR is very easy to GLOBAL_B 20.0
use. Indeed, it is the MR package used by Rupp’s MOLECULES 2
automated protocol (Rupp et al., (2002). It was successful SEED 357539
in finding the solution of 6PGL, using default options. SCALECELL 4.0
MAXGRIDSPACING 1.0
Use of Qs (Glykos and Kokkinidis, SCMODE wilson
2000, 2001, 2003) INTERPOLATION linear
Qs example.in >example.log & CELL 70.315 80.852 90.31 90. 90. 90.
where example.in will look like this: GROUP 19