PBC-related jumps in protein–protein contact analysis of a Martini CG simulation

User discussions
1

Hello everyone,

I am analyzing a Martini coarse-grained molecular dynamics simulation of a membrane-associated protein system containing three protein chains, A, B, and C.

I am specifically interested in protein–protein contacts between these chains. I am having difficulty determining whether some large jumps in the calculated minimum distance are genuine protein dissociation events or artifacts related to periodic boundary conditions (PBC).

The simulation is approximately 50 μs long, with 101 trajectory frames sampled every 0.5 μs.

For the A–B interaction, I calculate the minimum distance and number of contacts using:

gmx mindist -s production.tpr -f trj.xtc -n protein_index.ndx -od AB_mindist.xvg -on AB_contacts.xvg -d 0.6 -pbc yes

The A and B chains are separate index groups.

I obtain the following values around one of the problematic regions:

Time (μs) AB minimum distance (nm) AB contacts (<0.6 nm)

10.0 0.430 3
10.5 0.370 52
11.0 1.776 0
11.5 0.367 92
12.0 0.326 115
12.5 0.359 114
13.0 0.350 23
13.5 0.337 104
14.0 4.982 0
14.5 2.636 0
15.0 9.392 0
15.5 7.817 0
16.0 5.482 0
16.5 0.510 4
17.0 1.401 0

The A–B COM distance also shows similar large excursions.

I have also tried reconstructing the trajectory using:

gmx trjconv -s production.tpr -f trj.xtc -o whole.xtc -pbc mol -ur compact

and subsequently centering the trajectory.

However, the A–B distance still shows essentially the same large excursions after this processing.

I also tested the periodic-image distance for the individual protein chains using gmx mindist -pi. The protein-to-own-periodic-image distances remain approximately 6–9 nm and do not show a collapse around the problematic 14–16 μs region.

Therefore, I am unsure whether the large A–B distance excursions represent:

  1. genuine dissociation/reassociation of the two protein chains,
  2. A and B becoming separated across a periodic boundary,
  3. an issue with how the individual protein chains are reconstructed under PBC, or
  4. another issue related to analyzing a Martini CG protein complex.

My main question

For a Martini CG protein–protein complex, what is the recommended workflow for validating PBC effects before interpreting gmx mindist/contact results?

In particular, should I:

  • create a combined A+B index group and use -pbc cluster,
  • use -pbc nojump with a properly reconstructed first/reference frame,
  • use another PBC reconstruction procedure,
  • or simply perform the contact analysis directly on the original trajectory with -pbc yes?

I am mainly concerned about the validity of the quantitative contact analysis.

I would particularly appreciate advice from anyone who has experience with Martini coarse-grained protein–protein complexes and PBC-related contact/distance analysis.

Thank you
ishwar