AlphaFold prediction of complex - results not downloaded

Ladies and Gentlemen, This weekend, I did an AlphaFold prediction run of the human Troponin complex (Troponin C, T and I, UniProt-Ids P63316, P45379, P19429, respectively) via a freshly installed ChimeraX (v 1.4)/Google Colab on macOS Catalina (10.15.7). It completed successfully after 8 hrs (see screenshot below) but I failed to locate the results. According to your tutorial (https://www.youtube.com/watch?v=6lXeCPuTePs and https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html), I was expecting the results in a Prediction#-folder within the Downloads folder, but ChimeraX had only saved the UniProt xml files there. Also, the notebook code indicated the download of a "results.zip" folder but that also cannot be found anywhere on the computer. What have I done wrong? Would I have to edit the notebook? Specify a folder, and where? And: can I still recover the results from Google Colab servers? I copied the notebook to my GoogleDrive but was unable to identify a run-id or anything specific to my run. I would really like to get this working as the prediction of complexes is such a fantastic resource! Many thanks for your help, Erik Werner, Berlin /--// //Dr. Erik Werner// //Paulstr. 36// //10557 Berlin// //Germany/

Hi Erik, Sorry, our AlphaFold / Colab expert is away this week, but I will try to answer. As I understand it, the structure(s) should appear in ChimeraX automatically after a successful calculation. That they did not open automatically suggests there was some other problem. In addition, there should be a directory under ~/Downloads/ChimeraX/AlphaFold/ containing the results, as mentioned in the tutorial you were viewing: <https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html> I found this one post about saving intermediate results when there is a calculation failure... not sure if it pertains to your situation, however: <https://www.rbvi.ucsf.edu/pipermail/chimerax-users/2022-July/004041.html> If I come up with something more definitive, will let you know. Maybe other users who regularly use AlphaFold via Colab can comment? Best, Elaine ----- Elaine C. Meng, Ph.D. UCSF Chimera(X) team Department of Pharmaceutical Chemistry University of California, San Francisco
On Aug 15, 2022, at 12:51 AM, Erik Werner via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ladies and Gentlemen,
This weekend, I did an AlphaFold prediction run of the human Troponin complex (Troponin C, T and I, UniProt-Ids P63316, P45379, P19429, respectively) via a freshly installed ChimeraX (v 1.4)/Google Colab on macOS Catalina (10.15.7). It completed successfully after 8 hrs (see screenshot below) but I failed to locate the results.
<Screenshot%202022-08-14%20at%2021.48.30.png>
According to your tutorial (https://www.youtube.com/watch?v=6lXeCPuTePs and https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html), I was expecting the results in a Prediction#-folder within the Downloads folder, but ChimeraX had only saved the UniProt xml files there. Also, the notebook code indicated the download of a "results.zip" folder but that also cannot be found anywhere on the computer.
What have I done wrong? Would I have to edit the notebook? Specify a folder, and where? And: can I still recover the results from Google Colab servers? I copied the notebook to my GoogleDrive but was unable to identify a run-id or anything specific to my run.
I would really like to get this working as the prediction of complexes is such a fantastic resource!
Many thanks for your help, Erik Werner

Hi, Elaine, I am not an expert, but had the same problem, solved by trial and error. It involves using alpha-fold independetly from Chimera, and uploading the result in Chimera X. I will be interested if there is a more direct way. Claudio On Mon, 15 Aug 2022 at 17:52, Elaine Meng via ChimeraX-users < chimerax-users@cgl.ucsf.edu> wrote:
Hi Erik, Sorry, our AlphaFold / Colab expert is away this week, but I will try to answer.
As I understand it, the structure(s) should appear in ChimeraX automatically after a successful calculation. That they did not open automatically suggests there was some other problem. In addition, there should be a directory under ~/Downloads/ChimeraX/AlphaFold/ containing the results, as mentioned in the tutorial you were viewing: < https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html
I found this one post about saving intermediate results when there is a calculation failure... not sure if it pertains to your situation, however: <https://www.rbvi.ucsf.edu/pipermail/chimerax-users/2022-July/004041.html>
If I come up with something more definitive, will let you know.
Maybe other users who regularly use AlphaFold via Colab can comment? Best, Elaine ----- Elaine C. Meng, Ph.D. UCSF Chimera(X) team Department of Pharmaceutical Chemistry University of California, San Francisco
On Aug 15, 2022, at 12:51 AM, Erik Werner via ChimeraX-users < chimerax-users@cgl.ucsf.edu> wrote:
Ladies and Gentlemen,
This weekend, I did an AlphaFold prediction run of the human Troponin complex (Troponin C, T and I, UniProt-Ids P63316, P45379, P19429, respectively) via a freshly installed ChimeraX (v 1.4)/Google Colab on macOS Catalina (10.15.7). It completed successfully after 8 hrs (see screenshot below) but I failed to locate the results.
<Screenshot%202022-08-14%20at%2021.48.30.png>
According to your tutorial (https://www.youtube.com/watch?v=6lXeCPuTePs and https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html), I was expecting the results in a Prediction#-folder within the Downloads folder, but ChimeraX had only saved the UniProt xml files there. Also, the notebook code indicated the download of a "results.zip" folder but that also cannot be found anywhere on the computer.
What have I done wrong? Would I have to edit the notebook? Specify a folder, and where? And: can I still recover the results from Google Colab servers? I copied the notebook to my GoogleDrive but was unable to identify a run-id or anything specific to my run.
I would really like to get this working as the prediction of complexes is such a fantastic resource!
Many thanks for your help, Erik Werner
_______________________________________________ ChimeraX-users mailing list ChimeraX-users@cgl.ucsf.edu Manage subscription: https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users

Hi Erik, The screen shot of the AlphaFold prediction panel does not show that the job finished. Your screen shot shows it is calculating model 4. That means it has completed 3 model calculations. But AlphaFold by default computes 5 models using 5 neural nets that were trained differently. It scores the 5 models and usually you look at the best of those models. None of the models are downloaded until all 5 are computed. So my guess is that your job did not finish. Possibly Google Colab, the remote server that the calculation runs on, timed out. Google Colab is meant to be used interactively and so they can disconnect you if your computer goes to sleep for some period of time (30 minutes? 1 hour?). Google does not document or explain any rules about when they well disconnect your session. Unfortunately there are no alternative free services that provide GPU computation, so that is why ChimeraX uses the unreliable Google Colab servers. Your 700 amino acid complex should complete in much less than 8 hours. I would expect 1 hour or less using current ChimeraX which uses ColabFold, an optimized version of AlphaFold. I would suggest running it again and set your computer not to sleep if you will be a way from the computer. Tom
On Aug 15, 2022, at 12:51 AM, Erik Werner via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:
Ladies and Gentlemen,
This weekend, I did an AlphaFold prediction run of the human Troponin complex (Troponin C, T and I, UniProt-Ids P63316, P45379, P19429, respectively) via a freshly installed ChimeraX (v 1.4)/Google Colab on macOS Catalina (10.15.7). It completed successfully after 8 hrs (see screenshot below) but I failed to locate the results.
<Screenshot%202022-08-14%20at%2021.48.30.png>
According to your tutorial (https://www.youtube.com/watch?v=6lXeCPuTePs <https://www.youtube.com/watch?v=6lXeCPuTePs> and https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html <https://www.rbvi.ucsf.edu/chimerax/data/alphafold-dec2021/af_multimer.html>), I was expecting the results in a Prediction#-folder within the Downloads folder, but ChimeraX had only saved the UniProt xml files there. Also, the notebook code indicated the download of a "results.zip" folder but that also cannot be found anywhere on the computer.
What have I done wrong? Would I have to edit the notebook? Specify a folder, and where? And: can I still recover the results from Google Colab servers? I copied the notebook to my GoogleDrive but was unable to identify a run-id or anything specific to my run.
I would really like to get this working as the prediction of complexes is such a fantastic resource!
Many thanks for your help,
Erik Werner, Berlin
participants (4)
-
Elaine Meng
-
Erik Werner
-
Scazzocchio Claudio
-
Tom Goddard