AF2 using MMseqs multimer model ranking error?

[jfbazan]

A long-running 3-way multimer job had finally reached the end of its prediction run, and was set to do the ranking step (& lightly refine the top model), when the program frustratingly crashed with the following error msg below. This is the first time I've run into this, but I'm unsure about re-launching the whole enchilada if the job will crash again, hours down the line. Thx for any clarity & help, cheers, FB

2023-12-03 17:20:01,939 reranking models by 'multimer' metric

ModuleNotFoundError Traceback (most recent call last) in <cell line: 65>() 63 64 download_alphafold_params(model_type, Path(".")) ---> 65 results = run( 66 queries=queries, 67 result_dir=result_dir,

5 frames /content/alphafold/relax/cleanup.py in 20 import io 21 ---> 22 import pdbfixer 23 from openmm import app 24 from openmm.app import element

ModuleNotFoundError: No module named 'pdbfixer'

---

NOTE: If your import is failing due to a missing package, you can manually install dependencies using either !pip or !apt.

To view examples of installing some common dependencies, click the "Open Examples" button below.

[milot-mirdita]

Seems like openmm/amber was not installed correctly. Did you setup colabfold with the localcolabfold installier?

I recently added colabfold_relax as a standalone script for relaxing. Ranking the predictions should be relatively straightforward based on the scores file which was hopefully generated.

[jfbazan]

Hi Milot, this was with the regular ColabFold, not locally installed. And yes, I use the standalone Relax Colab daily, very useful. However, the reasons for this particular crash are mysterious, and I’m running a shorter test job as we speak to see if it recurs! Thx kindly for your help, cheers, FernandoSent from my iPhoneOn Dec 3, 2023, at 12:03 PM, Milot Mirdita @.***> wrote: Seems like openmm/amber was not installed correctly. Did you setup colabfold with the localcolabfold installier? I recently added colabfold_relax as a standalone script for relaxing. Ranking the predictions should be relatively straightforward based on the scores file which was hopefully generated.

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: @.***>

[HosseinEbrahimiK]

Same issue here! Running colabFold on google colab.