524
S. von Grafenstein et al.
is a degree of freedom in the docking algorithm (see Sect. 15.2.1.4). In both cases
the best suitable receptor conformation for each ligand is selected individually.
Selecting the best pose from multiple hit lists is for example used in a study using
ensembles of X-ray structures and is also referred to as “merging and shrinking
procedure” [70]. For various targets it was shown that the enrichment of active
ligands is improved in comparison to the use of a single X-ray structure [114]. The
merging strategy was used to compare the performance of representatives derived
from standard MD simulation and replica exchange MD simulations to X-ray structures of three systems [57]. Applying the hit list ranking by best pose, MD-derived
ensemble representatives perform at least similar to X-ray structures. This indicates,
that MD-derived ensemble docking will be useful in cases where alternative X-ray
structures are not available [57]. Xu et al. compare the effect of selecting the most
favorable score over an averaging of multiple scores which will be discussed below
[115].
No difference is made if the best score is a single event calculated on a receptor conformation occurring only once in the whole trajectory. Taking experimentally derived structural ensembles the composition of the available structures has
no physical meaning. In some cases, there is experimental evidence that a structure
representing a specific conformation should be favored or penalized. For example,
the formation of a subpocket might be an endothermic process and the energy of
pocket formation is needed to be compensated by ligand binding [116]. Attempts
to include the energy of individual conformations in scoring similar to the conformational energy for a ligand pose are made for ensemble docking with X-ray structures. Wei et al. show that the integration of a conformational strain for the receptor
conformation improved the enrichment of the ensemble docking [116]. They used a
docking algorithm where flexible parts are sampled from alternative conformations
during docking for a study on T4 lysozyme and thymidylate synthase. The energy
for deformation of the binding pocket in reference to the ligand-free conformation
was estimated by a special function for cavity formation within T4 lysozyme [117]
or an approximation via the change in total electrostatics and the change in accessible non-polar surface area for thymidylate synthase. Also Barril and Morley showed
that a penalty term for the formation of an alternative conformation elevated the
number of ligands docked with RMSD lower 2 Å compared to the native binding
pose using a set of 149 heat shock protein 90 structures [118].
Inclusion of a conformational strain for the receptor conformation is not common
for MD-derived ensembles. Assuming converged sampling, the distribution of the
conformations within the MD ensemble will reflect not only the strain enthalpy of
individual conformers but also the free energy. Conformations with unfavorable energy will be visited less often in comparison to conformations of lower free energy.
This might be the physical background why averaging over multiple scoring values
is frequently used to combine scoring values from docking to MD ensembles.
The relaxed-complex-scheme implements sampling information in a straightforward way by taking the average of scoring values of a ligand when docked to the
individual periodically extracted conformations. The developers of RCS observed
in a first application that the most favorable scoring values resulted from non-native
S. von Grafenstein et al.
is a degree of freedom in the docking algorithm (see Sect. 15.2.1.4). In both cases
the best suitable receptor conformation for each ligand is selected individually.
Selecting the best pose from multiple hit lists is for example used in a study using
ensembles of X-ray structures and is also referred to as “merging and shrinking
procedure” [70]. For various targets it was shown that the enrichment of active
ligands is improved in comparison to the use of a single X-ray structure [114]. The
merging strategy was used to compare the performance of representatives derived
from standard MD simulation and replica exchange MD simulations to X-ray structures of three systems [57]. Applying the hit list ranking by best pose, MD-derived
ensemble representatives perform at least similar to X-ray structures. This indicates,
that MD-derived ensemble docking will be useful in cases where alternative X-ray
structures are not available [57]. Xu et al. compare the effect of selecting the most
favorable score over an averaging of multiple scores which will be discussed below
[115].
No difference is made if the best score is a single event calculated on a receptor conformation occurring only once in the whole trajectory. Taking experimentally derived structural ensembles the composition of the available structures has
no physical meaning. In some cases, there is experimental evidence that a structure
representing a specific conformation should be favored or penalized. For example,
the formation of a subpocket might be an endothermic process and the energy of
pocket formation is needed to be compensated by ligand binding [116]. Attempts
to include the energy of individual conformations in scoring similar to the conformational energy for a ligand pose are made for ensemble docking with X-ray structures. Wei et al. show that the integration of a conformational strain for the receptor
conformation improved the enrichment of the ensemble docking [116]. They used a
docking algorithm where flexible parts are sampled from alternative conformations
during docking for a study on T4 lysozyme and thymidylate synthase. The energy
for deformation of the binding pocket in reference to the ligand-free conformation
was estimated by a special function for cavity formation within T4 lysozyme [117]
or an approximation via the change in total electrostatics and the change in accessible non-polar surface area for thymidylate synthase. Also Barril and Morley showed
that a penalty term for the formation of an alternative conformation elevated the
number of ligands docked with RMSD lower 2 Å compared to the native binding
pose using a set of 149 heat shock protein 90 structures [118].
Inclusion of a conformational strain for the receptor conformation is not common
for MD-derived ensembles. Assuming converged sampling, the distribution of the
conformations within the MD ensemble will reflect not only the strain enthalpy of
individual conformers but also the free energy. Conformations with unfavorable energy will be visited less often in comparison to conformations of lower free energy.
This might be the physical background why averaging over multiple scoring values
is frequently used to combine scoring values from docking to MD ensembles.
The relaxed-complex-scheme implements sampling information in a straightforward way by taking the average of scoring values of a ligand when docked to the
individual periodically extracted conformations. The developers of RCS observed
in a first application that the most favorable scoring values resulted from non-native