I got empty files when I tried to use the "1 - Segmentation of a TPS structure into TPS-specific domains"

[Daydream0513]

I ran the example you provided, and the log file output is as follows:

All residues: 0%| | 0/8 [00:00<?, ?it/s]
All residues: 12%|█▎ | 1/8 [00:00<00:00, 7.41it/s]
All residues: 75%|███████▌ | 6/8 [00:00<00:00, 27.56it/s]
All residues: 100%|██████████| 8/8 [00:00<00:00, 28.08it/s]

0it [00:00, ?it/s]
0it [00:00, ?it/s]

0it [00:00, ?it/s]
0it [00:00, ?it/s]

These files in "detected domains" are all empty.

I tried to solve the problem but got no job.

[SamusRam]

Dear @Daydream0513 ,

Sorry to hear about your difficulties. Let's make the tool work for you.

I assume you used the default domain TPS templates, or did you try some domain templates of your own?

If we're talking about TPS, let's first ensure that your structures have TPS domains. Could you please upload a structure you used here
and run the analysis (just hit Run all and then upload any of the structures)? This colab notebook uses my installation of the tool at our server as a backend; if there are TPS domains, it should tell us.

P.S. Btw, as a temporary workaround, feel free to use the API's following the code from the referenced colab, and run domain detection on my server.

[Daydream0513]

I guess it is due to the lack of secondary structure information in my structure file.

These TPS structures were predicted from Alphafold3 with a CIF file format and subsequently converted to a PDB file format using Pymol.

I'm trying to fix this problem for the secondary structure. Do you think this problem is a key factor in my failures? Or do you have any other better suggestions?

I'll tell you whether my job is working.

[SamusRam]

Dear @Daydream0513,

I've implemented conversion from a CIF format to PDB format directly in the Colab notebook

I've tested it on Alphafold3-produced CIF files, and it should function as expected.
Could you please try processing your files via the Colab and let me know the result? Just upload your AF3-generated CIF file there.

P.S. During the effort, I noticed that my structure-processing functions might get confused by non-alphanumeric characters in the filenames. I address it explicitly in the Colab. In your local pipelines, it's safer to avoid underscores in file names.

[Daydream0513]

It works fine in Colab!

I noticed that you updated an issue, so I reinstalled the tool manually and avoided special characters.

The good news is that the example you provided works correctly, but it reports an error when I enter the structure I provided.

Traceback (most recent call last):
File "XXX/lib/python3.10/runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "XXX/lib/python3.10/runpy.py", line 86, in _run_code
exec(code, run_globals)
File "XXX/src/structure_processing/domain_detections.py", line 492, in <module>
cmd.select(
File "XXX/lib/python3.10/site-packages/pymol/selecting.py", line 85, in select
return _cmd.select(
pymol.CmdException: Error: Malformed selection.

I tried to solve this problem but was unsuccessful, and it isn't an input issue. I used the colab you provided me earlier, and my file was successfully divided into domains.

And I found that the file path in the code for extracting the domain structure reported an error once, possibly because we applied the Slurm system.

[SamusRam]

Hey @Daydream0513 ,

I'm sorry to hear that you're still experiencing issues with the local installation.
I'm glad to hear that the colab works on your inputs.

And thanks for sharing the details. The error indicates a PyMOL issue with residue subselection.
How many structures do you need to proceed? In the Colab, you can check how I call the API from my server and use it programmatically on your end.

Meanwhile, could you share the problematic input structure? I can look into it and try to replicate the issue on my end.