Disrupted protein structure after 5 nanoseconds of GROMACS molecular dynamics

Question

[Nestor Wendt]

Hello, I have been trying perform a molecular dynamics using GROMACS. The idea is to simulate a system with the Human FABP2 and palmitic acid. My simulation is based on this tutorial. After about 5 nanoseconds my structure starts to show a strange behavior, as shows the figure.

Starting structure (left). Structure after 5 nanoseconds of simulation (right).

Does anyone know what causes this problem? I basically followed the tutorial, using the GROMOS96 force field. I suspect that this kind of behavior is called 'blowing up', although I am not sure and want to hear the opinion from more experienced bioinformaticians.

Thank you

[glihm]

Did you use the same parameters as tutorial, or did you adapt them to you simulation? I personally use Amber package, and I have never observed this king of 'blowing up'.

What are your parameters? (Time step, pressure, Temperature, cut-off?) Thanks to provide details. ;)

[Nestor Wendt]

I used the same parameters. The only change was in the ions, as my structure had a different net charge if compared to the protein used in the tutorial. I did not see this behavior in the equilibration steps, only in the final dynamics. A few parameters:

• Time step = 2 fs
• Pressure = 1 bar
• Temperature = 300 K for all groups

I am not sure if that's the cut-off parameters you are looking for, but the cut-off scheme was Verlet.

If I run the simulation without the ligand, using another force field, this disruption does not happen. I have been trying to find a tool to generate the ligand topology for another force field, but I could not find a working one.

Answer 1

[Nestor Wendt]

For those who encounter this problem in the future. After converting the trajectory file (.trr) to (.xtc) using gmx trjconv with the following parameter -pbc nojump, these jumps stop appearing. This tutorial called this jumps 'pdc artifacts', and there, before visual analysis they were removed. Here's the gmx trjconv help page.

Thanks.