Closed xiki-tempula closed 2 years ago
Could you please send all the relevant files?
So the file for reproducing the xTB topology is in folder xTB
.
I don't have any QM software yet so I did a hacky way where I generated the QM hessian with psi4 (QM/lig_hessian.npy
) and use the xTB routine to generate the new topology, I think you could regenerate the QM topology with any other QM engine. The topology file that I have generate is in QM/gas.top
.
100 conformers are in MM/lig.gro
and MM/traj.xtc
, I used run.sh
to generate the MM energy profile and plot.py
to analysis the result.
energy.pdf
As is shown here that the QM one is similar to other force field and seems ok but the xTB one is very wrong.
Archive.zip
Thanks.
; flexible dihedrals 15 17 18 20 3 -14476.443 2997.159 1898.391 -752.774 -1962.776 -2373.229
Something seem to have gone wrong with one of the dihedral fitting. You should probably have received a warning about this.
@selimsami Yes (scan_data_CN_H12C10N2O1_75efeccb473e9b543b61dc50f5d1fc16), it seems a bit weird that the same dihedral in the on the other one is fine where the only thing that have changed is the hessian.
@selimsami Wait let me do some further investigation on this.
Starting geometry of the MM optimization is based on the QM optimized fragment. The difference between those is where I would start the investigation
Somehow I cannot reproduce the issue now.
I'm testing the xTB routine and for most molecules (>95%). The xTB hessian behaves fine (give similar energy to PDB0-D3(BJ)/def2-TZVP). However there is one molecule that gives vastly different energy.
The way that I test the force field is that I generate 100 conformations of the molecule, and compare the MM energy of these 100 conformations.
The error seems to come from the angle part. However, I cannot diagnose the issue, I wonder if you mind help me on this? Thanks. The molecule is