Hi Vincent,

  AlphaFold uses OpenMM to energy minimize structures.  If you really want to understand the details then look at the AlphaFold code.  In run_alphafold.py 

https://github.com/google-deepmind/alphafold/blob/7c9114c8423ac9db981d8365168464bab09b3e54/run_alphafold.py#L527

you can see some of the parameters it uses

    amber_relaxer = relax.AmberRelaxation(
        max_iterations=RELAX_MAX_ITERATIONS,
        tolerance=RELAX_ENERGY_TOLERANCE,
        stiffness=RELAX_STIFFNESS,
        exclude_residues=RELAX_EXCLUDE_RESIDUES,
        max_outer_iterations=RELAX_MAX_OUTER_ITERATIONS,
        use_gpu=FLAGS.use_gpu_relax)

and the details are in the alphafold/relax code

https://github.com/google-deepmind/alphafold/tree/main/alphafold/relax

Energy minimization is a common way for AlphaFold predictions to fail.  Sometimes that happens because a bad clash in the prediction cannot be resolved.  Because the minimization is prone to failure and tends to hide areas where the AlphaFold prediction had problems, ChimeraX has energy minimization off by default.

Tom

On Oct 1, 2023, at 12:53 PM, Elaine Meng via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:

Hi Vincent,
The AlphaFold calculation is separate from ChimeraX, and we are not the developers of that program.  

For details on how it works, I can only suggest reading the AlphaFold, AlphaFold-Multimer, and perhaps also the ColabFold papers cited in our help page.  There are links to all three papers therein:

AlphaFold and AlphaFold-Multimer paper links near the top:
<https://rbvi.ucsf.edu/chimerax/docs/user/tools/alphafold.html#top>

...and ColabFold paper link in the prediction section:
<https://rbvi.ucsf.edu/chimerax/docs/user/tools/alphafold.html#predict>

If these papers are not providing adequate detail, you could consider asking the AlphaFold developers (authors of the papers) directly.

Best,
Elaine
-----
Elaine C. Meng, Ph.D.                       
UCSF Chimera(X) team
Department of Pharmaceutical Chemistry
University of California, San Francisco

On Oct 1, 2023, at 10:39 AM, Vincent Candela via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:

Hello,

I am an undergraduate researcher at McGill University using the Alphafold feature of ChimeraX to study the role of tau on microtubule dimers, and I was wondering if you could point me to detailed documentation on the energy-minimization feature of Alphafold. I’ve been looking at the ChimeraX documentation and couldn’t find an in-depth documentation on what it does.

Thank you so much,

Vincent Candela


_______________________________________________
ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu
To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu
Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/