Hi everyone,
I am trying to do molecular dynamics simulation of carbondioxide gases flowing (permemeate) through lipids membrane such as DPPC (Dipalmitoylphosphatidylcholine). I am using coarse grained martini force field. I am having difficulity to combine the DPPC data that uses martini force field with carbondioxide gases that uses TraPPE force field. I am using lammps and moltemplate.
Does anyone has experiences making similar simulation?
Hello,
One thing that you can do is coarsen your CO2 into one bead and assign it a bead flavor within MARTINI.
Another thing that you can do (although not exactly the most proper way to do it) is to keep the CO2 in the atomistic resolution, the membrane in the coarse grained and use mixing rules for the force-fields between (CO2 atoms)-(membrane beads). This is not really something that people usually do I think, so probably you will be “inventing” this way of modelling your system. But you can give it a shot and see if you are able to reproduce whatever it is that you are interested at with this strategy. I myself once did something similar when I wanted to investigate something and I didnt have time to derive force-fields in a more intelligent way 
Also beware that if you are interested in dynamical properties (such as diffusion coefficient), it s very much possible that you will have a problem if you are gonna keep the gas in atomistic and the membrane CG’d. It is quite pssible that the dynamics of the gas will not be stable at high timesteps. If it is, there is still the matter if your model is indeed “properly reproducing the dynamics” since the timestep of MARTINI is too aggressive for the atomistic. Moreover, you would have the problem of having to figure out how much time the timestep actually corresponds to as in CG it is often the case that the timestep loses the meaning of actual time.
1 Like
It is generally not considered a good idea to mix and match empirical force fields and that is particularly true to do this when you want to combine all-atom and coarse grain force fields. The problem is that either force field has a different “philosophy” in how to derive parameters and specifically how to balance between Lennard-Jones and Coulomb components of the interactions. If those strategies don’t match, the mixed terms are likely bogus and lead to inconsistent results. Thus, you now either have to look up how to model CO2 with Martini, your lipid with TRAPPE, or find a different suitable general force field than has atom types for both components. OPLS-AA would be a strong candidate for that. Or CHARMM/CGenFF.
1 Like
If you want to keep everything in the CG scale due to the computational cost and derive force-fields with actual physical meaning, you can also check this software called VOTCA. It allows you to use several different strategies (such as force-matching, iterative boltzmann inversion, relative entropy minimization). You can study these methods and see which one derives force-fields following a methodlogy that you would expect to ultimately better reproduce whatever it is you are interested at.
PS: there is this paper (https://pubs.acs.org/doi/10.1021/jp044629q) in which force-matching was used to derive potentials for a system similar to yours. VOTCA implements force-matching with the same formalism as the one in Sergei and Voth’s paper except the for the Coulombic interaction term, which is not implemented in VOTCA’s original version (there is a branched version of it in which it is implemented though). In any case, it might be a good bet
1 Like
Thankyou very much for your comprehensive answer!, it’s really helpful
I am also considering the first options, converting the CO2 into one bead and incorporate it to the Martini force field, but i can’t figure out how to convert the TraPPE parameter for CO2 into a coarse-grained parameter.
I will try using VOTCA package that your recommend, i will update the results later, thanks! 
Thankyou very much for your advice!, i think i will try to find out how to model CO2 with martini as you suggest, since it is more easier
Well, in my suggestion related to modelling CO2 with MARTINI you would dismiss entirely the TraPPe force-field. The idea would be to classify CO2 with a given bead type (e.g., C3) and the parameters for CO2-CO2 and CO2-(membrane beads) would come from the MARTINI parametrization.