Conformational Entropy Changes

[ioanaapapa]

Conformational entropy results for the small molecules are now significantly higher - seem unusually high and no longer match the SAMPL paper ones. The difference happened sometime after CodeEntropy1.0.3 and before the conformational entropy averaging PR, which did not affect this issue.
For G1, one of the guest molecules

[jimboid]

Thanks for reporting @ioanaapapa.

What would really helpful, because there are 3 changes that might have caused this, is either can you drop your parameters for how you ran this example to allow us to reproduce for 1.0.3 and then run some tests against the other versions. Or you could create separate conda envs with 1.0.4, 1.0.5 and see which version the issue start with which would tell us which of the changes brought about the difference.

[ioanaapapa]

These are the log files for when I ran this with 1.0.3:

program_com.txt

program_output.txt

program_log.txt

And these are the topology and trajectory for just the ligand:

G1_traj.zip

[jimboid]

Thanks. Will have a look

[harryswift01]

@ioanaapapa @jimboid @skfegan

I’ve now had the opportunity to run four versions of CodeEntropy using identical inputs, driven by the same config.yaml file.

---

run1:
  top_traj_file: ["../../test_data/G1_traj/G1_stripped.tpr", "../../test_data/G1_traj/G1_stripped.trr"]
  selection_string: 'resname LIG'
  temperature: 300

From an initial comparison, the first noticeable difference appears between v1.0.3 and v1.0.4. Specifically, the conformational entropy for group 0 at the united‑atom level changes from:

• 21.338046088170707 in v1.0.3
• to 49.94564885738359 in v1.0.4

The latter value is then unchanged in subsequent releases.

CodeEntropy version	Conformational entropy
v1.0.3	21.338046088170707
v1.0.4	49.94564885738359
v1.0.5	49.94564885738359
v1.0.6	49.94564885738359

The change is introduced in v1.0.4, which corresponds to the v1.0.4 release tag. One potential explanation for this behaviour may be changes introduced in PR #167, which appears to be the most relevant modification in that release that could affect conformational entropy calculations.

I've also attached all of the final results for each version if you would like to look into those with more detail.

v1.0.3-output_file.json
v1.0.4-output_file.json
v1.0.5-output_file.json
v1.0.6-output_file.json

[ioanaapapa]

It looks like the conformational entropy doesn't change with the number of frames.

[ioanaapapa]

Closed by PR #267.