510
S. von Grafenstein et al.
ligand-free structure. All glutamic acid residues in the binding site are deprotonated.
Although there are attempts to include reprotonation events in MD simulations with
constant pH simulations [51], classical force fields do not allow an adaption of the
protonation state. With the assignment of atom types in simulation setup, protonation and tautomer states of amino acids are determined for the simulation. The
effect of protonation on docking is essential as the interaction potentials, especially
electrostatic interactions, will differ for a deprotonated or protonated glutamate.
Recently, two studies addressed the aspect of binding site protonation for rigid molecular docking [52, 53]. Especially docking results for binding sites containing a
histidine residue were shown to be dependent on protonation and tautomer states, as
for this amino acid various alternative states can occur [53]. The altered interactions
of a protonated versus deprotonated amino acid will also influence the conformational sampling during the MD simulation.
15.2.1.3 Alternative Sampling Methods
The conformational transitions observed between X-ray structures in the ligand–
free and ligand-bound state can include loop shifts of several Ångströms or the
reorientation of whole protein domains [42, 55]. The time scale of such events is
typically not covered statistically by all-atom simulations ranging in the area of
several dozens of nanoseconds [12]. Focusing on these transitions several groups
established different procedures to improve the sampling in comparison to classical unbiased MD simulation. We refer the interested readers to a recent review
on enhanced sampling techniques in the context of drug design and the references
therein [56]. Here we will only focus on methods used in connection with molecular
docking.
We group methods which rely on the same principles as MD simulation, e.g.,
torsional angle MD [47], replica exchange MD (REMD) [57, 58] and targeted MD
[59]. Also other enhanced MD sampling techniques, such as accelerated MD
could be applied [56]. Torsional angle MD is not an enhanced sampling technique
in the classical sense as no bias is applied to overcome energy barriers more easily. Sampling using torsional angle MD is more efficient than standard MD as it
takes advantage of an internal coordinate system based on the relative orientations
along the dihedral angles in a protein [47]. However, torsional angle MD cannot
be applied to systems with multiple molecular entities such as solute plus solvent
molecules hindering its broad application.
In REMD techniques higher temperatures allow an enhanced sampling [60].
Multiple copies of a system are simulated at different temperatures and an exchange
between the replicas is regulated by a Metropolis criterion [61]. The technique allows to overcome energy barriers more easily and was shown to be suitable to reproduce a ligand-bound-like conformation of a flexible RNA target [58]. Osgurthorpe
et al. compared the performance of docking to REMD-derived versus classical
MD-derived ensembles. However, they could not observe an improvement on the
discrimination between active and inactive ligands for the subsequent docking [57].
S. von Grafenstein et al.
ligand-free structure. All glutamic acid residues in the binding site are deprotonated.
Although there are attempts to include reprotonation events in MD simulations with
constant pH simulations [51], classical force fields do not allow an adaption of the
protonation state. With the assignment of atom types in simulation setup, protonation and tautomer states of amino acids are determined for the simulation. The
effect of protonation on docking is essential as the interaction potentials, especially
electrostatic interactions, will differ for a deprotonated or protonated glutamate.
Recently, two studies addressed the aspect of binding site protonation for rigid molecular docking [52, 53]. Especially docking results for binding sites containing a
histidine residue were shown to be dependent on protonation and tautomer states, as
for this amino acid various alternative states can occur [53]. The altered interactions
of a protonated versus deprotonated amino acid will also influence the conformational sampling during the MD simulation.
15.2.1.3 Alternative Sampling Methods
The conformational transitions observed between X-ray structures in the ligand–
free and ligand-bound state can include loop shifts of several Ångströms or the
reorientation of whole protein domains [42, 55]. The time scale of such events is
typically not covered statistically by all-atom simulations ranging in the area of
several dozens of nanoseconds [12]. Focusing on these transitions several groups
established different procedures to improve the sampling in comparison to classical unbiased MD simulation. We refer the interested readers to a recent review
on enhanced sampling techniques in the context of drug design and the references
therein [56]. Here we will only focus on methods used in connection with molecular
docking.
We group methods which rely on the same principles as MD simulation, e.g.,
torsional angle MD [47], replica exchange MD (REMD) [57, 58] and targeted MD
[59]. Also other enhanced MD sampling techniques, such as accelerated MD
could be applied [56]. Torsional angle MD is not an enhanced sampling technique
in the classical sense as no bias is applied to overcome energy barriers more easily. Sampling using torsional angle MD is more efficient than standard MD as it
takes advantage of an internal coordinate system based on the relative orientations
along the dihedral angles in a protein [47]. However, torsional angle MD cannot
be applied to systems with multiple molecular entities such as solute plus solvent
molecules hindering its broad application.
In REMD techniques higher temperatures allow an enhanced sampling [60].
Multiple copies of a system are simulated at different temperatures and an exchange
between the replicas is regulated by a Metropolis criterion [61]. The technique allows to overcome energy barriers more easily and was shown to be suitable to reproduce a ligand-bound-like conformation of a flexible RNA target [58]. Osgurthorpe
et al. compared the performance of docking to REMD-derived versus classical
MD-derived ensembles. However, they could not observe an improvement on the
discrimination between active and inactive ligands for the subsequent docking [57].