Assessment of Common Simulation Protocols for Simulations of Nanopores, Membrane Proteins, and Channels.

Molecular dynamics (MD) simulation has become a common technique to study biological systems. Transport of small molecules through carbon nanotubes and membrane proteins has been an intensely studied topic, and MD simulations have been able to provide valuable predictions, many of which have later been experimentally proven. Simulations of such systems pose challenges, and unexpected problems in commonly used protocols and methods have been found in the past few years. The two main reasons why some were not found before are that most of these newly discovered errors do not lead to unstable simulations. Furthermore, some of them manifest themselves only after relatively long simulation times. We assessed the reliability of the most common simulations protocols by MD and stochastic dynamics (SD) or Langevin dynamics, simulations of an alpha hemolysin nanochannel embedded in a palmitoyloleoylphosphatidylcholine (POPC) lipid bilayer. Our findings are that (a) reaction field electrostatics should not be used in simulations of such systems, (b) local thermostats should be preferred over global ones since the latter may lead to an unphysical temperature distribution, (c) neighbor lists should be updated at all time steps, and (d) charge groups should be used with care and never in conjunction with reaction field electrostatics.