Atom model
All-atom model in Molecular Dynamics can be achieved in CHARMM, Amber, NAMD, GROMACS, LAMMPS packages and so on. Among different force fields, CHARMM force field is the most popularly used one to study biomolecular interactions. From my own experience, I use GROMACS,Amber and NAMD a lot to deal with nanoparticles, large polymers, protein-protein interaction and ion channels. For each package, there are always advantages and disadvantages to simulate MD system. For example, high performance clusters with GPU can speed up systems running in Amber, but in terms of CPU, it seems GROMACS and NAMD are much faster than Amber and CHARMM.
Amber tools:
Pytraj and Cpptraj are powerful analysis tools for trajectories running in Amber.
PARMED is useful in selecting groups and generating different topologies.
MDAnalysis is also useful in analyzing Amber trajectories, especially the function LeafletFinder()
MDStress is very useful for local stress calculations and also analyze GROMACS trajectories .trr format. And cpptraj can also change Amber trajectories to GROMACS trajectories with velocities.
VMD can also do a lot of analysis for all-atom systems.
Coarse-grain (CG) Model
Martini force field is a coarse-grain representation of four-to-one mapping (i.e. on average four heavy atoms and associated hydrogens are represented by a single interaction center) which preserves chemical properties and reduces computational cost. CG Martini Force field can do excellent job in speeding up protein and bilayer membrane system simulation. The simulation can easily achieve up to time scales in the micro- to millisecond range; if using GPU driven clusters. Although CG model loses atom level information such as protein folding, the backward-mapping engine from Martini group give an opportunity to go back to all-atom model, based on which we can implement atomic simulation and analysis. Overall, CG modeling has several powerful pros, 1) capability of efficient simulations of huge system sizes; 2) allowing for simulation of slow processes by speeding-up; 3) systematically exploring state conditions in thousands of parallel runs; 4) providing a computationally inexpensive testing ground for exploring novel generic biophysical pathways[Ref].
Martini tools:
Insane.py is used for membrane packing tool
Martinize.py is to make CG protein structure topologies from atom structures
Backward.py is to map CG model back to all-atom configuration, (CHARMM-GUI: Martini Maker option can also do reverse-mapping using Martini Backward-mapping algorithm)
MDAnalysis is a toolbox designed for MD analysis, especially helpful for protein and bilayer membrane system. Have to combine with jupyter notbook in python.
PyCGTOOL is a method that can quickly and easily generate coarse-grained molecular dynamics model within MARTINI framework from atom simulation. At the same time, this can also help validate CG model and provide details for benchmark.
MDStress is very useful for local stress calculations from molecular simulations. It is based on double precise GROMACS version.
Tips:
When changing trajectories or pdb file from different package, we have to remember that histidine is recognized as HIS in GROMACS, but as HSD in CHARMM force field. It is tricky and can cause problem about pdb reading, and may give errors like missing patches in protein.
Insane.py gives the gro file when building up the bilayers, but we should be careful, since the last line in .gro file will have some non-zero values at the last 6 columns, which will make the simulation have non-90 degrees pbc. This non-90 degree pbc will make domain decompositions in cluster fail. Be careful, since I got this error feedback when using XSEDE in COMET and STAMPEDE2.
It is good to use psf2itp.py provided in CHARMM-GUI tools, to change CHARMM psf to gromacs topology. This can be used for many proteins not in GROMACS pdb2gmx format, and it is quick and easy. Usage: ./psf2itp toppar.str protein.psf
When we truncate a system and want to rebuild it, it is better to have 0.5A more (e.g. 2*0.5 A longer in x-axis) on each side for pbc box to have additional space for system relaxation and re-equilibration, or it may give bad conformation and the system will crash immediately.
Amber adding voltage: If efn is on (efn=1), the x, y, z (efx, efy, efz) components are scaled to box size. For example efx/x length of box size, efy/y length of box size, efz/z length of box size. This normalizes the electric field charge to your box size. It is off when it is 0. It currently only supports pmemd (both the serial and MPI versions). we have .200 eV * (1 e) * 23.061 (kcal/mol) / eV * (1 / 150 Ang) = .030748 kcal/(mol * e * Ang) . [efn=1,efx=0,efy=0,efz=-4.612]
If using CG model and then reverse-mapping, try to have the larger position restraint and elastic network restraint to make sure the secondary structure of protein is not twisted.