Biomedical Engineering Reference
In-Depth Information
indicate participation in an interface. A simple approach is to identify residues
with values that are higher than a threshold, such as the mean CSP plus one
standard deviation.
166
An alternative procedure is to reject all residues with
values higher than the mean CSP plus two standard deviations, then re-
calculate the mean and standard deviation, and repeat the process iteratively
until no residues are rejected; finally, all of the rejected residues are classed as
being significantly perturbed.
166
Both methods are prone to highlighting false
positive sites and neglecting false negatives. Most residues that participate in
interfaces possess high solvent accessibility in the unbound state, whereas CSPs
from buried sites are usually the result of subtle conformational changes
associated with complex formation; hence, it is customary to exclude residues
with low relative solvent accessibility (e.g., ,50%).
167
On the other hand,
[
1
H,
15
N]-HSQC spectra lack signals from proline resides, so these sites may be
overlooked even if they genuinely participate in protein-protein interactions.
The SAMPLEX program has therefore been developed to consider the
structure of a protein together with a list of experimental CSPs, returninga
per-residue list of confidence values that can distinguish between perturbed
and unperturbed regions.
166
A cautious strategy for interpreting sites with significant CSP values is
adopted by the HADDOCK algorithm, which converts them into ambiguous
interaction restraints (AIRs) to be used both for steering docking simulations
and for assessing the quality of models of the complex.
167,168
Surface sites on
the partner molecules are classed as active residues (known to form
intermolecular contacts in the complex), passive residues (which potentially
make contacts), or inactive residues (not involved in the interface). An AIR
restraint is defined between a single active residue on one protein and every
active or passive residue on the partner molecule;
167
this forces putative
interface residues to come within 3-4
˚
of a surface region on the partner
protein, without dictating specific contacts between any particular active or
passive sites.
168
The docking protocol comprises three stages: a rigid body
energy minimisation (to sample compatible poses); a semi-flexible refinement
(to allow backbone and side-chain atoms near the interface to rearrange); and
a final refinement in explicit solvent (to optimise packing and improve the
energetics of the interface).
167
The structures obtained in the final round of
calculations are clustered using the pairwise backbone RMSD of residues at
the interface (i-RMSD, defined for all residues within 10
˚
of the partner
molecule) and ranked according to a composite energy score. In favourable
test cases, the lowest energy complexes possessed i-RMSD values ,2.0
˚
from
reference structures.
167
The HADDOCK data-driven docking approach has
been highly popular, leading to the deposition of coordinates in the PDB for
.60 complexes containing up to six components, although some of these make
use of mutagenesis, hydrogen/deuterium exchange or sparse NOE restraintsin
addition to or instead of CSP data.
169
Weak points of the method concern the
reliability and completeness of the experimental data set, problems discrimi-
nating the relative orientation of partners at the interface, and difficulties with
