Torsion scan skipped in ethanol example despite rotalltors=True in poltype.ini

[KeitoHASEGAWA]

Hi, I’m currently testing Poltype2 using the ethanol example, and I’ve encountered an issue where torsion scans are not being performed even though I have set rotalltors=True in poltype.ini.

Here are some relevant details:

I'm using the official ethanol example structure (ethanol.sdf) and my poltype.ini contains the following:

structure=ethanol.sdf
numproc=8
maxmem=16GB
maxdisk=50GB
# fit all torsion parameters
rotalltors=True

The job completes, but the torsion scan steps are entirely skipped. Specifically, no qm-torsion/ directory is generated under the Temp/ folder.

Below is a snippet from my poltype.log (note: no actual torsion job submissions):

Mon May 26 15:05:39 2025 Submitting: python /home/.../lAssignAMOEBAplusPRM.py -xyz ethanol.xyz_2 -key ethanol_prefitvdw.key -sdf ethanol.sdf -potent TORSION path = /.../Temp
Mon May 26 15:05:40 2025 Poltype Job Finished

I attached the log file

For reference, here is a snippet from a working log (provided by the developer):

Wed Feb 19 10:38:40 2025 Relaxed Torsion Scan and Fitting
Wed Feb 19 10:38:40 2025 Submitting: minimize.x ethanol-opt-1-2-060_preQMOPTprefit.xyz ...

I also checked rotalltors parsing in poltype.py, and found this line:

elif "rotalltors" in newline:
    self.rotalltors = self.SetDefaultBool(line, a, True)

This suggests it should be properly recognized. The .ini file is definitely being read, since the structure and other resources are processed as expected.

Is there any reason why rotalltors=True would be silently ignored?
I downloaded the Poltype2 source files on May 14, 2025 from the GitHub repository
Python 3.9.13
OS: Linux (RHEL8-based), kernel: 4.18.0-372.9.1.el8.x86_64
(https://github.com/TinkerTools/poltype2) as a ZIP file (not via git clone), so the repository metadata (e.g. commit hashes) are not available.

ethanol-poltype.log

[KeitoHASEGAWA]

Follow-up: Unexpected Torsion Parameter Assignment Despite rotalltors=True

rotalltors=True
I confirmed that at the initial stage, rotalltors=True works as expected: torsionsmissing.txt is correctly populated with all rotatable dihedrals.

However, later during the generation of ethanol_prefitvdw.key_torsion, the script appears to assign torsion parameters from the database — for example:

# Matched by lAssignAMOEBAplusPRM.py: Ether prop O-CH2-CH3-H from amoeba09
torsion 401 402 403 405 0.000 0.0 1  0.000 180.0 2  0.300 0.0 3

This causes:

The original torsionsmissing.txt to be renamed to .back

A new, empty torsionsmissing.txt to be created

→ Therefore, no torsion QM fitting is performed

This contradicts the rotalltors=True intention and likely indicates a bug or unintended behavior in the handling of the rotalltors flag during the key_torsion generation step.
I will attach ethanol_prefitvdw.key_torsion and torsionsmissing.txt.back for reference.

Thank you for your continued support!

2files.zip

[leucinw]

fixed in this commit

[coffee-on-tap]

For awareness, this doesn't work for cyclic (tested on non-aromatic, like PC) compounds. Step 26 seems to match the ring torsions for benzene which is related to another issue. A cursory glance at the old database parsing code seems to imply that rings are always assumed to be aromatic as a kind of workaround. Perhaps this is why benzene params get matched. Given that you're also migrating to a different parser script, you might end up addressing some of these issues by eliminating the legacy parser entirely.

Is there any interest in later developments leveraging more established tools for torsion scans?

[KeitoHASEGAWA]

Yes, the problem that this doesn't work for cyclic (tested on non-aromatic, like PC) compounds sees to be another issue. I'm intereested in the later development for solving this problem.

[leucinw]

For awareness, this doesn't work for cyclic (tested on non-aromatic, like PC) compounds. Step 26 seems to match the ring torsions for benzene which is related to another issue. A cursory glance at the old database parsing code seems to imply that rings are always assumed to be aromatic as a kind of workaround. Perhaps this is why benzene params get matched. Given that you're also migrating to a different parser script, you might end up addressing some of these issues by eliminating the legacy parser entirely.

PC is working properly when setting (rotalltors=True). It only fit the C-C rotatable bond and other torsions are assigned with the latest commit.

[leucinw]

As well, ethanol is working as excepted when set rotalltors=True with the latest commit.