Closed bernieleecy closed 3 years ago
Hi there. Thanks for posting.
You shouldn't need to manually convert the water topology, this will be done for you on write by BioSimSpace.IO.saveMolecules
.
For your AMBER simulations that are unfolding, are you using sander
or pmemd
? We recently discovered and fixed an issue with zero periodicity torsions that need to be treated differently in pmemd
. In our own simulations we also experienced protein unfolding. See here for details. Could you report what version of Sire you are using, so we know that you have the fixed applied.
import Sire
print(Sire.__version__)
For the difference in the bonded terms: It's not possible to directly compare AMBER and GROMACS bond energies due to the way in which they treat three-point water models. To do so, you'd need to make sure that they are both implementing SHAKE in the same way, which might mean tweaking the config file options. See this thread for a discussion.
I'll try to reproduce your numbers here when I get a chance.
I've been using pmemd
and Sire 2020.1.0. I've also been using the stable release of BioSimSpace rather than the latest development version, so that might be why I needed to fix the waters manually.
I'll download the development version and test it and let you know if there are any further issues.
Thanks for the update. You'll need the latest version of Sire, 2021.1.0. Let me know if this doesn't work for you and I'll look into it further.
Another think that I thought of is that AMBER can have issues if molecules in the topology are wrapped across the periodic image. I've not checked your topology, but it might be worth having a look if it still doesn't work with the most recent version of Sire. MDAnalysis has functionality to unwrap the coordinates if needed.
Cheers
This also reminds me that I need to push out a new release of BioSimSpace sometime soon!
It's fine with Sire 2021.1.0, I ran a 10 ns simulation and the protein was stable.
Thank you for your help!
No problem, glad to hear that it's now working. This was a really peculiar issue that caused us a lot of headaches to track down and solve.
Thanks again for reporting!
I tried to convert a solvated and minimised Gromacs topology to Amber format using BioSimSpace, with the water conversion fix described in #52 applied.
These files were then used for simulations in Amber 20, i.e. BSS was used to convert the topology only. I equilibrated the system (NVT then NPT, total of 600 ps of restrained equilibration), then ran a short unrestrained MD simulation (1 ns). In the unrestrained MD simulation, the protein was unstable (started to unfold), and the 1-4 EEL value looks odd. The energy values at the end of the simulation were:
I went back and checked the energies of the starting topologies. For the unconverted topology, the energies were as follows (from gmx mdrun -rerun):
For the converted topology, the energies were as follows (from sander). The main difference I could see was in the bonded terms, rather than the electrostatic terms.
I also tried to convert the Amber topology back to Gromacs format, and checked the energies again. The energies were very similar to the original Gromacs topology.
From this, I'm not sure what's causing the instability in MD simulations. I originally thought it was the dihedrals, but the single point calculations on the minimised structure suggest that's not the case?
key_files_bss.zip