I'm intending to perform a simulation of the transport of simple ions or water through perm-selective membranes, using Molecular Dynamics (MD) and/or Monte Carlo (MC). I've read around that, in general, with Monte Carlo the polymerization of the membrane can be modeled starting from linear chains and monomers, and get the pseudo-membrane well cross-linked. Once the polymer structure is optimized, MD is performed with explicit solvent, the membrane putted in between two different solutions, so transport properties through membranes can be explored.

I never used MD nor MC. This is where I stand in the knowledge:

Monte Carlo (stochastic approach): Useful to calculate thermodynamic macroscopic properties. The simulation follows the Markov Process, where the steps are independent one from each other (Jan Labanowski).

The stochastic approach, called Monte Carlo, is based on exploring the energy surface by randomly probing the geometry of the molecular system. The iterations are independent of one another (i.e., the system does not contain any "memory''). Is accepted more by physicists than by chemists, probably because MC is not a deterministic method and does not offer time evolution of the system in a form suitable for viewing. (But) for Markov chains, there are efficient methods for deriving time related quantities such as relaxation times (…) Many chemical problems in statistical mechanics are approached more efficiently with MC, and some (e.g., simulations of polymers chains on a lattice) can only be done efficiently with MC.

Molecular dynamics (deterministic approach): To find trajectories and get an insightful movie of the molecular movement, with parameters of interest. Normally the system composes of around 5000 atoms with explicit solvent (water molecules generally). Periodic conditions are employed. The Verlet algorithm and the SHAKE routine are commonly used; here the bond lengths are fixed and the whole molecules are allow to move between certain potential range given (in position and energy), this repeated successively through a minimum path over the potential surface. It's a leap-frog method.

Considering this, I basically don't know where to begin. I would like to know the basic howto's and softwares for MD (GROMACS, CHARMM, LAMMPS, AMBER, etc.) and ways to perform Monte Carlo for polymers. Also, besides the fundamentals, I would like to know if there is like a Handbook for the features of each MD software or modes of application (such us potentials/ functionals used, force fields), or the most important practical keys to be taken into account before simulating.

Thanks in advance,

  • 4
    $\begingroup$ I would recommend "Understanding Molecular Simulation [amazon.com/Understanding-Molecular-Simulation-Second-Edition/dp/… ]" by Frenkel and Smit for the theory part $\endgroup$ – mamun Dec 3 '15 at 18:52
  • 2
    $\begingroup$ Another enlightening book is "Computer Simulation of Liquids", by Allen and Tildesley. $\endgroup$ – user23061 Dec 4 '15 at 14:51
  • 1
    $\begingroup$ Hi @Orr22, your link seems broken (in the comment above), could you fix that? $\endgroup$ – Felipe S. S. Schneider Jan 20 '17 at 1:12
  • $\begingroup$ Great thanks for the alert, finally I found this as good introduction amzn.to/2jK5RuS. But, and, especially the documentation of each molecular dynamics software, they are very self-explanatory. $\endgroup$ – Orr22 Jan 21 '17 at 14:03
  • 1
    $\begingroup$ en.wikipedia.org/wiki/MARTINI $\endgroup$ – Mithoron May 13 '17 at 20:04

I always find AMBER to be the most streamlined for conducting MD, however it (as well as CHARMM) are proprietary. Note that both are Langevin dynamics based, whereas LAMMPS is brownian dynamics.

That being said, an MD simulation is only as good as the parameters for the force field; the engine is of no matter (unless we're discussing speed). Basically, I would use a QM program (Gaussian is my goto, but nwchem or psi4 are excellent open source options) to optimize the geometry of the membrane. Because this is rather heavy, you'd most likely be best to go with a semi-empircal method as ab-initio could take weeks (depending on your computational power). Once optimized, use the QM software to derive force constants as well as ESP charges (partial charges) for each atom. Take this data, and parameterize a force field with it (I'm partial to using the Amber FF's...google GAFF or FF99SB... Theres a paper out there from Cornell, Merz, et. al. that should give you insight).

I hope this helps give you an idea of what you're looking at for a respectable MD simulation.

  • $\begingroup$ It really guide me through :) $\endgroup$ – Orr22 Jan 5 '16 at 12:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.