The BMW-MARTINI Coarse-grained Force Field
Please refer to the following references for detailed descriptions of the force field:
- The BMW water model (Z. Wu, Q. Cui, A. Yethiraj; J. Phys. Chem. B 2010, 114, 10524-10529)
- The BMW-MARTINI force field (Z. Wu, Q. Cui, A. Yethiraj; J. Chem. Theory Comput. 2011, 7, 3793-3802; with supporting info.)
- Difference between Arg and Lys (Z. Wu, Q. Cui, A. Yethiraj; J. Phys. Chem. B 2013, 117, 12145-12156)
Parameter files for the BMW-MARTINI force field
All parameter files are currently built for simulations with Gromacs. (Note: The version must be higher than 4.0.) But once the interaction tables are built, simulations can also be performed with CHARMM, NAMD, and so on.
The table files for different interactions:
Must use "user" defined VDW interaction in your mdp files (see the sample mdp files below). With the tables implemented, we use different cut-off schemes for different pair interactions (see supporting information in the force field paper for details.) Please define the corresponding types (i.e., SOL for water, NON for nonsolvent) in your index file and put them in the energygrps, and energygrp_table sections of your mdp files in order to use them.
The itp files for Gromacs simulations:
Please try NOT to mix these files with the original MARTINI force field itp files for your simulation. The interactions between the charged groups have been reparameterized; moreover, many other parameters have been adjusted, e.g., the angular term for lipid tails. If you need to generate your itp (or use itp files from the original MARTINI) for other chemical components, please be aware that: 1. type AC1 and AC2 in MARTINI are now replaced by C1 and C2; 2. RQd is used for for Arginine guanidinium group and KQd for Lysine, and AQa is representing Aspartate and Glutamate side-chains; 3. for anions (Cl-) and peptide C-ter, the MARTINI Qa is replaced by AQa. An example for making such changes can be found in the cgbax.itp attached in the next section.
Sample inputs and initial configurations
An example for minimizing and equilibriating a coarse-grained DOPC lipid bilayer with Gromacs (4.5.4) using the current BMW-MARTINI force field: example_dopc.tar.
BMW water box: bmw.gro
Can be used with Gromacs genbox to create your solvated system.
mdp file for membrane simulation: bmw_membrane.mdp
The time step is generally 20 fs. But it can be varied from 40 to even 5 fs depending on the specific system studied. Time step out of this range is not recommended.
Please try NOT to change the md setups such as parameters for neighbor searching, electrostatics and VDW.
Similar to the original MARTINI, using mdrun option -rdd value as 1.4 nm or even 1.6 nm might help if the system is not very stable.
During the optimization phase of a simulation, there might be problems with using Settle on the BMW water (as in atomistic simulations). Such problem can be circumvented by first relaxing the system with "flexible" BMW water (with harmonic bonding interactions between the three particles), then optimizing the system again with rigid water using Settle. To do this, first modify the martini_v2.1_bmw.itp file in the SOL water section, uncomment the bonding section and comment out the Settle part to make water "flexible", optimize the system, then change the file back and optimize the system again. (The above example_dopc.tar is a good example for doing this.)
top file for membrane simulation: bmw_membrane.top
itp file for helix 5 in Bax protein (Bcl-2 family): cgbax.itp
This is an example for converting the itp file for amino acids from the original MARTINI force field to the BMW-MARTINI format (changing the particle types according to instructions above). Please compare the above cgbax.itp file with the one from MARTINI (cgbax_mar.itp) and the modifications should be obvious (using vimdiff in Linux). Regarding how to generate peptide/protein gro and itp files, please see here and here, generate the MARTINI files first and change the types accordingly for the BMW model. In the current version of coarse-grained force field, peptide secondary structures are still not flexible and better models are being developed.
DPPC bilayers (128 lipids & 1971 CG water, 325K, equilibriated for 1000 ns): dppc.gro
DOPC bilayers (128 lipids & 1971 CG water, 300K, equilibriated for 1000 ns): dopc.gro
DOPE bilayers (128 lipids & 1100 CG water, 300K, equilibriated for 480 ns): dope.gro
DOPS bilayers (128 lipids, 128 Na+, & 1100 CG water, 300K, equilibriated for 480 ns): dops.gro, dops.top
To generate lipids with different head group, for example, DOPG, just simply replace DOPS with DOPG in the dops.top file. Then equilibriate the system (> 200 ns) with your new dops.top file, with "grompp -c dops.gro -f bmw_membrane.mdp -p dops.top -n -o -maxwarn 1". Now Gromacs will give you a warnning message about different lipid head group names but will use DOPG parameters in the simulation. The equilibriated configuration will be the one for DOPG.
For Charmm users, one might also want to check out the Cui group repository for MARTINI. Some inputs scripts (top, par and pdb2crd converter files) is provided and they can be use to generate the initial configurations with Charmm. (Note: It is not fully funtional for MD simulations in Charmm yet.)
In the current form, the model is not accurate for the mechanical properties of the membrane, with higher values for the area compressibility modulus and line tension compared to experiment and the original MARTINI model. We attribute this to the mapping of four water molecules to one CG unit which makes the process of transferring a water from the aqueous phase to hydrophobic region unrealistic. This overestimation of mechanical properties is likely to result in a high barrier for membrane pore formation and lipid flip-flop. An improved version is being developed.