Now that the temperature is stable, we must equilibrate with respect to pressure. The NPT phase for a membrane protein system is generally somewhat longer than for a simple aqueous protein, again due to the heterogeneity of the system; while water will diffuse rapidly, lipid diffusion is comparatively slow. Here, we will conduct a 1-ns NPT equilibration. The .mdp file can be found here.

There are a few changes in this .mdp file worth noting:

• tcoupl = Nose-Hoover: The Nosé-Hoover thermostat is widely used in membrane simulations because it produces a correct kinetic ensemble and allows for fluctuations that produce more natural dynamics. The Nosé-Hoover thermostat is not appropriate for NVT equilibration, since it allows for such wide fluctuations. Begin with Berendsen or V-rescale, and switch to Nosé-Hoover at the beginning of NPT or production MD.
• pcoupltype = semiisotropic: Uniform pressure scaling (isotropic) is not appropriate for membranes. A bilayer should be allowed to deform in the x-y plane independently of the z-axis. One can also use anisotropic pressure coupling for membranes, but be aware that substantial skewing of the box vectors can occur over very long simulation timeframes.
• There are now two values specified for both compressibility and ref_p, corresponding to values for the x-y and z dimensions, respectively.

Now, proceed with grompp and mdrun, as usual. Since the simulation will be 1 ns in length, it is best to run it in parallel on a cluster. Note that GROMACS 4.5 introduced threading for parallelization, meaning that on a multi-core workstation, an external MPI library is not required. For network-connected clusters, MPI is still needed for inter-node communication.

gmx grompp -f npt.mdp -c nvt.gro -r nvt.gro -t nvt.cpt -p topol.top -n index.ndx -o npt.tpr

gmx mdrun -deffnm npt

Analyze the pressure progression, again using gmx energy. It is also advisable to verify that the box vectors have stabilized, ensuring a stable lateral area of the membrane (Box-X and Box-Y).