Chemistry Reference
In-Depth Information
receive most of the computational effort. The use of empirical parameters is
reduced. Mostly, force-field description of the molecular system is required for
estimating binding free energies [587].
In the approach known as λ-dynamics simulation, there is the propagation of the λ
parameter with the atomic coordinates during the simulation process. The λ
parameter is treated as a dynamic variable with a fictitious mass. Multiple ligands
can be evaluated simultaneously. In the generalized ensemble TI method relative
free energies can be obtained from a single λ-dynamics simulation. Trajectory
snapshots can be sorted into bins and traditional TI equations used to compute
changes in free energies [584, 585].
Constant pHMD is a variation of the λ-dynamics method, with the usage of
fictitious λ particles to propagate titration degrees of freedom. In this method the
end-states represent the protonated (λ=1) and deprotonated (λ=0) states. It is
possible to incorporate in adiabatic free energy dynamics (λ-AFED) a λ
parameter. Free energy profiles along reaction paths can be generated. Swithching
functions generate high barriers between endpoints. Temperature and mass of λ
are assigned [630, 631].
Multiple ligands that do not interact with each other are represented in multiple
topology λ-dynamics representation. Biasing potentials are used as reference free
energies. Restraining potentials added so that ligands do not go outside the
binding pocket. The system partitioned into environment and individual ligands.
The λ is treated as a dynamic variable linked to Monte Carlo variations after steps
of molecular dynamics. The λ is also used as a self-regulating sampling variable
to explore barriers and low-energy regions. It is also possible to associate λ values
with multiple copies in order to sample regions of high energy. The λ-dynamics,
however, have often focused on simulating difference at single sites in a given
system [622].
In the path-integral method, a stochastic kinetic formalism is used to predict
relative binding affinities of protein-ligand complexes. The ligand is modeled as a
Brownian particle. The latter is subjected to nonbonding interaction potentials of
the receptor allowing determination of relative binding affinities. Near the binding
