Biology Reference
In-Depth Information
is to design a protein that binds a target protein of interest, starting from a structural model
(preferably a crystal structure) of the target protein alone. The computational workflow used
to tackle this problem can roughly be divided into three steps:
1.
Choosing a
protein, i.e. a protein that can serve as the backbone template that
the binding sequence can be designed onto.
2.
Finding a productive relative spatial orientation of the target protein and the scaffold.
3.
Designing the amino acid sequence of the scaffold (and potentially the target) to stabi-
lize this spatial orientation.
scaffold
'
'
The first two steps represent computational challenges unique to this problem. In step 1, the
scaffold protein can either be taken from a library of existing proteins for which the crystal
structure is known, or it can be designed de novo using category 2 CPD algorithms. In step 2,
after a scaffold has been decided upon, category 3 (rigid-body placement) algorithms then
need to be used to place it in a spatial orientation towards the target, thus giving an initial
model of the complex. Essentially, the
shape of the complex and the location of the
binding interfaces on the two partners are determined in step 2. This model is then passed to
side-chain placement algorithms (or to the specialized algorithms developed for protein
interface design as described above) to design the novel amino acid sequence.
'
global
'
Arguably, step 2 is the most critical stage in this workflow, because it is necessary to find a
spatial orientation that is
i.e. where a sequence can be designed such that the
resulting binding interface has sufficient size, shape complementarity, and interactions for
the complex to be energetically favorable compared to the unbound state and thus lead to a
high-affinity interaction. Moreover, many applications require that a particular region of the
target protein is part of the binding interface, meaning the rigid-body placement algorithm
needs to be able to orient the scaffold to optimally interact with the desired surface patch
on the target. In recent years, several impressive examples of successful de novo protein
interface design have been reported, using different approaches for steps 1 and 2. These
different approaches can be divided into two groups: general approaches that can in
principle be used to design a binder for any arbitrary target protein; and more specialized
approaches that take advantage of certain structural properties of the target.
designable,
'
'
107
Perhaps the most general method for scaffold selection and placement is Fleishman et al.
s
'
method,
23
shown in
Figure 6.1
, which was recently used to design
proteins that bind to a conserved region of influenza hemagglutinin and inhibit this
protein
so-called
Hotspot-design
'
'
s ability to undergo conformational changes that underlie influenza infectiousness.
24
This method is based on the observation that in many natural protein
'
protein interfaces,
the affinity is mostly mediated by a small subset of the residues making up the interface.
When making a series of single-point mutants of a binding protein, in which residues
belonging to the interface have been mutated to alanine, and measuring the resulting
binding affinities, usually only a subset of variants will have significantly reduced affinities.
The residues that were replaced by alanine in these variants are thus the most important
interface residues, and are often referred to as
residues.
25
From this observation,
Fleishman et al. reasoned that a viable strategy to design novel interfaces would be to first
place a small number of side-chains in favorable locations on the target interface, and then
use these proto-hotspot residues as anchors that the rest of the interface is designed around.
In the Hotspot-design method, a small (usually on the order of 1
hotspot
'
'
3) set of disembodied
amino acids is placed in the vicinity of the target-protein surface patch in an energetically
favorable fashion. For example, if there are exposed hydrophobics on the target surface, an
aromatic amino acid might be chosen as the hotspot-residue, or in case unsatisfied
hydrogen-bonding atoms are present, an amino acid with complementary hydrogen-
bonding functionality could be picked. The exact location for each hotspot residue can
either be set explicitly or found with the help of ligand-docking approaches.
26
Once the
hotspot residues have been placed, a candidate scaffold protein is placed such that the
