[gmx-users] Umbrella sampling question

[Gmx QA]

Hi

I'm a new poster on the maillist, and new to umbrella sampling but not to
MD in general.

I have recently done some pulling simulation with Gromacs, but have a few
questions about the outcome, and in particular the distances that are
calculated for the different windows. I realize this is a question that has
come up before, and there are some useful posts in the archives, but
nothing that exactly answers my questions.

To begin, I ran a pulling-simulation pulling a molecule starting in the
middle of a membrane (and inside a protein transport channel) straight up
in z. To that end, I used the following mdp-settings:

pull = umbrella
pull_geometry = direction
pull_vec1 = 0 0 1
pull_ngroups = 1
pull_start = yes
pull_group0 = Protein
pull_group1 = pg1
pull_rate1 = 0.001
pull_k1 = 500

So my pull groups are the COM of the protein, and one atom named pg1 on the
pull molecule.
To my understanding, this should pull the pg1-molecule straight up in z,
and that is indeed also what is happening in the simulation.

I ran this simulation for 6 ns, and it resulted in about 40 separate
conformations to use for umbrella sampling. All of those simulations also
seem to work, I can use the resulting pullf,xvg-files as input to g_wham,
and get a histogram-plot with reasonable overlaps.

However, I'm trying to understand how the various distances relate to each
other. For example, in the profile.xvg-file I get a z-distance for the
first frame as 1.761971 nm, while checking with g_dist gives me 1.80833
(for z). Continuing a few frames further, there are still differences, and
they appear to be random.

How are the distances in the profile.xvg-file computed? The average of the
dZ column of the pullx-file for the first frame is 1.75787, which is sort
of close to 1.761971, but not quite (and also for the next frames there is
no closeness).

Sorry if this became a long mail, but I need to understand this in order to
be able to progress with my research.

Thanks