Possible issue with SemiBin 2.2.0 on fresh conda environment

[tnmquann]

Hi,
I would like to report a potential issue when running SemiBin2 v2.2.0 on my cluster (without a GPU).

After creating a fresh conda environment, SemiBin2 repeatedly becomes stuck at the binning stage. The log stops progressing after the following lines:

[2025-11-19 09:05:52] INFO: Start binning.
[2025-11-19 09:05:53] DEBUG: Running command: /sbin/ldconfig (full command line: `/sbin/ldconfig -p`)
[2025-11-19 09:05:58] DEBUG: Calculating depth matrix.
[2025-11-19 09:05:58] DEBUG: Number of edges in clustering graph: 279041
[2025-11-19 09:05:58] DEBUG: Running infomap with 30 processes...

No further progress occurs unless the process is interrupted manually.

When terminating the run with Ctrl+C, the traceback suggests the issue may occur during the Infomap-based clustering step:

INFO Start binning.
^CTraceback (most recent call last):
  File ".../SemiBin2", line 10, in <module>
    sys.exit(main2())
  File ".../site-packages/SemiBin/main.py", line 1625, in main2
    single_easy_binning(...)
  File ".../site-packages/SemiBin/main.py", line 1317, in single_easy_binning
    binning_short(**binning_kwargs)
  File ".../site-packages/SemiBin/main.py", line 1222, in binning_short
    cluster(...)
  File ".../site-packages/SemiBin/cluster.py", line 289, in cluster
    embedding, contig_labels = run_embed_infomap(...)
  File ".../site-packages/SemiBin/cluster.py", line 182, in run_embed_infomap
    result = run_infomap(...)
  File ".../site-packages/SemiBin/cluster.py", line 25, in run_infomap
    p.join()
  File ".../multiprocessing/pool.py", line 665, in join
    self._worker_handler.join()
  File ".../threading.py", line 1094, in join
    self._handle.join(timeout)
KeyboardInterrupt

The environment was created using:

conda create -y -n semibin -c conda-forge -c bioconda semibin=2.2.0

Installation completes successfully, but the runtime issue persists.

Additional observations:

• If I install SemiBin 2.1.0 in a fresh environment, it runs normally without any hanging.
• If I then update this 2.1.0 environment to 2.2.0, the hanging issue does not occur. The update only adds safetensors=0.6.2 and upgrades SemiBin to 2.2.0.
• I also have another environment that includes SemiBin 2.2.0 along with several other binners; in that environment, SemiBin 2.2.0 runs fine.

I think this issue may be related to a missing or mismatched dependency that is present in updated/mixed environments but missing in a clean installation of v2.2.0.

I’ve attached the YAML files for both environments:

1. A fresh new environment created with semibin=2.2.0 (where the issue occurs).
   semibin_220_failed.yml

2. Environment that includes SemiBin 2.2.0 along with several other binners (where the issue does not occur).
   semibin_220_success.yml

Please let me know if you need further details. I’m happy to help diagnose the problem.

[dialvarezs]

I can confirm this issue, and it seems that igraph is the problem.
Starting from a fresh environment with Semibin 2.2.0, and downgrading igraph to <1.0 solves the issue for me.

[luispedro]

Thanks. I can reproduce it when I run SemiBin from scratch, but, weirdly, when I try to isolate the behaviour, it goes away...

[luispedro]

This is an openmp/multiprocessing interaction issue. Previous versions of igraph did not use openmp (at least in in this context). I will rewrite the code to work around it

[luispedro]

SemiBin 2.2.1 will fix this (just released, it normally takes a few hours for conda to pick it up)