Rerun xtc file generates different potential energies compared to original data

User discussions
1

GROMACS version: 2024.2
GROMACS modification: No

I need to rerun the simulation trajectories using a different tpr file (with different lambda value for free energy calculation) in order to use mbar method. I found that for the same trajectory, even using the same tpr file as the original simulation, rerun from xtc file generates different potential energies, although we still see a strong correlation (Fig. 1). However, rerun from trr file gives almost the identical values (Fig. 2, BTW, I set a very large saving interval for trr file).

Another interesting observation is that rerun from xtc can generate almost identical potential difference (namely the dH data in the dhdl.xvg file) as the original simulation (Fig. 3).

I’m just wondering if the discrepancy in potential energy between original data and the rerun data is due to the precision issue for the xtc file. Thanks!

PS: the production run was performed using GPU, rerun was done by CPU.

Figure 1. Rerun from xtc
Figure 1. Rerun from xtc1120×840 21.8 KB

Figure 2. Rerun from trr
Figure 2. Rerun from trr1120×840 19.6 KB

Figure 3. Potential difference between two different lambda values using same configurations
Figure 3. Potential difference between two different lambda values using same configurations1120×840 14.3 KB

2

This looks to be within the range expected of rounding coordinates to 0.001 nm. I don’t understand what U1 and U2 are.

3

Hi, Hess, thank you for the confirmation. U2-U1 is the difference in potential energies calculated using two different Hamiltonian for the same set of configurations , namely dH in the dhdl.xvg file from free energy calculation.