Pulling simulation for two particles

User discussions
1

GROMACS version: 4.6.7
GROMACS modification: YES
Hi,

I am trying to do a simple umbrella sampling between two particles in a box using ONLY nonbonded interaction between them. The potential between two particles is defined as LJ 12-10 which is provided by a table for Gromacs and defined in the topology file.
The reason that I am doing because I want to test it for a bigger system (Protein-ligand with structure-based model not all atoms), and want to make sure it is working correctly, firstly for two particles.

However, I got the wrong result and PMF is not like 12-10 potential between them. The PMF after a specific window starts to be like umbrella potential and the count plot shows there are not enough overlaps between them.

PMF
PMF1265×1123 45.6 KB

not overlap between them.

The system is a structure-based model (C_{/alpha}) provided by smog (The only option is groamcs 4), and non-bonded parameters are added manually to the top file.

I did not apply any minimization and equilibration for this system, this is a native structure-based system.

;pull code
pull = umbrella
pull-geometry = distance
pull-dim = Y Y Y
pull-start = no
pull-ngroups = 1
pull_group0 = A
pull_group1 = B
pull-nstfout = 100
pull-rate1 = 0.0
pull-k1 =10000.0

2

Two questions that come to mind are: do you have some overlap between the histograms for each window, and did you set correct values of the force constant?

4

This looks quite reasonable now. There is often a tiny artifact at the end, for many reasons (from PBC to sparse sampling at the tail of the last window), so that’s something to ignore - just trim your PMF at 1.3. But you’re right there is no attractive region, which makes me question how the attractive 12-10 potential was implemented (AFAIK it’s not available in Gromacs, except maybe through Plumed).

What I don’t understand is that you first say it’s two particles in a box, and then say it’s a native structure-based C-alpha model. Which one is that in the end?