Problem in the umbrella histrograms

User discussions
1

GROMACS version: 2019.4
GROMACS modification: No

Dear all,

After pulling a polymer through a membrane using steered molecular dynamics simulation, I selected the frames of this trajectory using the pull-x.xvg output as the initial configurations for the umbrella sampling simulations. A part of the list of these 111 frames x COM distance (z axis) is shown below:

 1 10.768
  2 10.464
  3 10.202
  4 9.919
  5 9.587
  6 9.313
  7 8.984
  8 8.689
  9 8.411
 10 8.113

However, after running umbrella sampling simulation and the wham, I noticed that the first histogram (black one below) is not located (centralized) at 10.768 nm as shown in the list above. The umbrella histograms are shown below:

histogram-problem
histogram-problem988×765 10.6 KB

Should I correct the position of this histogram by increasing the force constant in the umbrella sampling simulation (I used k = 1000 kJ/mol/nm)? I also have many gaps in these histogram despite the fact that the COM distances of each configuration were equally spaced by 0.3 nm as shown in the list above ( a part of this list of 111 configurations). Can the number of bins used in the wham calculation correct this problem or it is a question of increase the force constant in the US simulations?

Sincerely,

Eduardo.

2

Your force constant is far too low.

Note that pulling a large molecule through a membrane is a very challenging calculation. You might not get convergence with the simulation times available.

3

Thank you for your response. If I increase the force constant in the US simulations, can the histogram gap problem get worse? Do you have a suggestion on what this increase should be?

4

The distributions move towards the center of the potential when increasing the force constant. But you will need many more umbrella. The width of the distributions is proportional to 1/sqrt(k).

5

Thank you! I will increase the force constant and the number of umbrella in the regions with gaps. Should the increase in the force constant be done in the pulling simulation (steered molecular dynamics) or in the umbrella sampling simulations?

6

Strictly speaking only in the US simulations. But also, or maybe even more, in a steered simulation there is the risk that the system will lag and be far from the center of the umbrella.