8 Notes
- In case of proteins undergoing significant conformational
changes upon binding, sampling of available conformational
space would be useful. In such cases, molecular dynamics can
be used as a source of additional conformations. However,
sufficient sampling would require computationally expensive
long simulations. Moreover, there would be no guarantee
that all obtained conformations would be gathered from one
well of potential, restrained by energetic barriers easy to over-
come after the other protein binding. Therefore, accelerated
molecular dynamics can be a tempting alternative—it is less
computationally expensive and can easily overcome energetic
barriers. However, it should still be handled with care, espe-
cially when applied to transmembrane proteins. Accelerated
MD of a transmembrane protein immersed in a simplified, to
fluid membrane may result in artifacts. At least, appropriate
content of cholesterol should be ensured. - PPD methods take into account the flexibility of protein
backbones and side chains to a very limited degree. As the
formation of oligomers can induce significant structural
changes in the participating molecules, it is prudent to refine
the obtained complexes using molecular dynamics. Impor-
tantly, in the case of membrane proteins the complex should
be embedded in a lipid bilayer of a similar composition as the
environment where the modeled interaction is occurring. In
the first step the simulated system should be equilibrated over
the course of around 20 ns in npt conditions, with restraints
applied to the backbone of the studied proteins. This step
ensures sufficient lipid packing around the studied complex.
In the step, the restraint should be gradually released from the
complex during 20 ns, to enable the proteins to structurally
adapt to each other and the environment. Finally, the complex
should be simulated without any restraints until the backbone
RMSD is converged. - A benchmark of the available PPD methods [16] has shown
that most of the methods generate a complex that is highly
similar to the experimentally solved structure. The pose is not
selected as the best solution, due to the employed scoring
methods. Thus when performing PPD of membrane proteins,
the scoring method should be carefully considered. The
method should not take into account parameters specific for
aqueous proteins, like desolvation energy.
300 Agnieszka A. Kaczor et al.