GROMACS version:2018.4
GROMACS modification: No
I have a question. What do you use most when you try to measure the width of the membrane. I used three methods:
For phospholipids I create electron density profiles for the phosphor group and check the distance, but I also hear that people use only phosphorus atoms, not phosphorus group, so which is better?
For glycolipids I create electron density for lipids and water and check in which place these two profiles cross and I measure a distance between them.
Is it ok?
This is more in the lipid simulation domain than a Gromacs question. In my experience the measurement used is arbitrary. Just try to be consistent, and report exactly how you measure. If you need to compare with another publication, use their measurement system to the closest approximation you can.
As Kevin writes, it is somewhat arbitrary. This is stuff I did a lot already as a PhD student (sadly a LONG time ago), and it gets even more complex for large membrane systems where you start to have undulations, i.e. “waves” that propagate/oscillate through the system.
If you want to define local thickness, one of the best options (IMHO) is to simply define a specific atom in the lipids and measure from that one. Ideally you should pick one that doesn’t fluctuate too much, such as the carbon connecting the head with the tails. This is easy to calculate and average, even close to a membrane protein, and it will have the advantage that the local thickness should be the same as the size of the system grows, since you don’t average over the entire system. However, it can have issues e.g. if you have mixtures of several different types of lipids (since there might not be obviously corresponding atoms), not to mention you won’t have any obvious experimental correspondence to the specific thickness.
Another alternative is to use g_density or a similar too to calculate the density of certain groups (e.g. the hydrophobic CH2, or the carbonyls, or even the density distribution for a single atom like above). This can have the advantage of being more well-defined, and the densities distributions might often be possible to compare to e.g. neutron scattering or similar data (meaning: there’s a direct experimental counterpart). However, the challenge you get is that the undulations in larger systems will smear out the density if you just calculate it relative to the Z coordinate, so the distribution can depend on system size if there is no surface tension making the membrane flat.
The simple solution to this is usually to go for relative comparisons, which leads to conclusions that echo Kevin’s:
If comparing to somebody else’s data, use their definition.
If you want to make an argument about differences in systems/setups, find a definition where you can reasonably argue it’s equivalent.