Biomedical Engineering Reference
In-Depth Information
nucleus types; most structural features can be reproduced successfully when
13 C a / 13 C b -only or 1 H N / 13 C a / 15 N-only shift sets are supplied. 141,142 More
serious problems emerge when missing shift assignments are clustered together,
as may occur in turn regions that experience line broadening due to sampling
of multiple conformations, in paramagnetic metalloproteins because of
proximity to unpaired electrons, or in solid-state NMR studies, when signals
from surface-exposed residues may be undetectable due to sample hetero-
geneity. A hybrid version of CS-ROSETTA which uses alternative modelling
procedures in extended regions that lack shift data is more successful in such
cases; this has been used to determine accurate models (C a RMSD within 1.0 ˚
of reference structures) from incomplete sets of solid-state 13 C a , 13 C b , 13 C9 and
15 N isotropic chemical shifts for ubiquitin and the streptococcal GB3
domain. 142 Similarly, when using solid-state data from 1 H a , 13 C a , 13 C b and
15 N nuclei for both GB1 and the Src SH3 domain, CHESHIRE returned
structures that had acceptable accuracy (RMSD ,1.4 ˚ ), even in regions that
displayed significant shift differences between the solid and liquid states. 143
Chemical shift-based structure prediction methods have a wide range of
potential applications. At the simplest level, CHESHIRE, CS-ROSETTA and
CS23D have been used to verify that structures of protein domains in solution
are unchanged from those determined by X-ray crystallography 144,145 or by
NMR experiments on larger constructs. 146,147 These methods offer simple
ways to confirm that mutated proteins 148 , close homologues of previously
studied systems 149-151 or in silico designed proteins 152 possess the expected
fold. As a component of more conventional structure determination efforts,
CS-ROSETTA results are of sufficient quality to provide initial phases for
molecular replacement in X-ray crystallography, 153,154 to facilitate the assign-
ment of NOESY peak lists in NMR studies, 155 or to cross-validate models
derived using other experimental restraints. 156 These tools also provide
unprecedented opportunities for obtaining near-atomic resolution structural
information about protein samples with short lifetimes, 157
sparsely populated
excited states and transiently formed intermediates. 158,159
CS-ROSETTA and CHESHIRE regularly fail to converge for medium-sized
globular proteins that contain more than 110 residues, perhaps due to
inadequate sampling of conformations that are close to the global energy
minimum. 136,137,139 Although improvements in model accuracy may be gained
by using more faithful chemical shift back-prediction routines, such as
SPARTA+ 113 and SHIFTX2, 111 or by taking side-chain shifts into
account, 111,125,126 it is not clear that these modifications will extend the size
limit to larger systems. One way forward is to use additional experimental data
as a filter to help select the best models. For small to medium-sized proteins
(,130 residues) with complete backbone and side-chain assignments available,
this extra information could be the unassigned raw peak list from a NOESY
experiment. 155 However, backbone assignment of large slow-tumbling proteins
typically requires deuteration at CH sites, so side-chain assignments may
not always be available. Incorporation of backbone
1 H N - 15 N RDC data into
Search WWH ::




Custom Search