Biomedical Engineering Reference
In-Depth Information
residues.
26
120
This
approach
has
also
been
extended
to
symmetric
oligomers.
60,111
In a different family of bottom-up approaches, protein backbone is
assembled from consecutive peptide planes whose relative orientations can
be determined from RDC data by solving algebraic equations with only 16
discrete solutions.
58,59,112,113
Accurate structures of test proteins like ubiquitin
and the B1 domain of Protein G have been determined with this approach
using 3-4 RDCs per peptide plane in two alignment media.
58,112,114
Recently,
the RDC-Panda approach combined this algebraic method with stochastic
sampling and has been shown to solve structures using only N-H and C-H
RDCs in a single alignment medium. To extend the backbone structure
accurately across loop-regions a few long-range NOEs are beneficial.
115
Structure determination of protein complexes is a problem-class where sparse
NMR data is historically most frequently applied. Usually, one follows a
bottom-up strategy where the respective protein domains are determined as
isolated sub-units that are subsequently assembled in a rigid-body docking. For
two sub-units this requires only the sampling of six degrees of freedom for
rotation and translation.
61,76,116-120
In some approaches, e.g., the CNS-based
HADDOCK
120
and RosettaDock,
129
side-chains in the interface region are
allowed to move during a subsequent refinement phase. Due to the restriction to
rigid-bodies during the docking phase, thorough sampling can be achieved. If,
however, the structures of the isolated domains cause clashes when moved into
the native relative orientation and distance, a rigid-body docking protocol will
fail. In such cases an incorporation of side-chain and backbone flexibility into
docking protocols
60,111
can be beneficial, although it is more challenging to
achieve thorough sampling. Instead, refinement of the isolated domains against
NMR data obtained from the complex is often sufficient.
60,119
Coordinate
inaccuracy of the isolated domains can have an equally detrimental impact as
conformational changes between bound and unbound structure. We found in
some cases that CS-ROSETTA structures of the isolated domains fare better in
docking simulations than NOE-based NMR solution structures.
60
4.4.3 Resolution-Adapted Structural Recombination
In the fragment-assembly approaches discussed above, the conformational
space searched in the sampling phase is strongly reduced, which greatly
facilitates de novo structure calculation. However, large proteins and high-
contact order topologies remain challenging.
121,122
In fact, structures solved
with chemical-shift-based fragment assembly have generally fewer than 100
residues with relatively low contact order.
22,23
Within the ROSETTA
framework it has been shown that additional sparse NMR data—RDCs,
and
backbone H
N
-H
N
contacts—can
guide
sampling
towards
the
native
structure, and thus push the limit to ca. 120-130 residues.
22,26
To overcome the size and fold-complexity limitations inherent to fragment
assembly, we have developed an iterative sampling protocol that recombines
