Hi Nadine, As Elaine says, AlphaFold will find lots of template structures from the PDB (up to 20), but then only uses 4. The rules AlphaFold uses for which 4 to choose are idiotic. First it excludes any templates with more than 95% identity (ie exclude the best templates!). Then it makes no effort to cover your sequence and I think just take highest sequence identity. Unfortunately Google DeepMind seemed to put no thought into how to use templates. But the good news is it hardly matters. If AlphaFold can make a confident prediction without templates it doesn't matter what template you give it, it simply gives you the prediction it likes. I think only in cases where AlphaFold struggles to find the right fold will be helped by templates. If templates are turned off then it uses no template. It does not need any template to run. If I recall correctly it does represent the structure and it sets all the residue positions to 0 initially when there are no templates. If you want to understand this you need to read the original AlphaFold papers. Tom

On Feb 6, 2024, at 11:02 AM, Elaine Meng via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:

Hi Nadine, AlphaFold is not a homology modeling program per se, so there may be less difference when using or not using templates than you might expect. By "not a homology modeling program" I mean it isn't strictly dependent on using specific template structures, but has instead learned general rules about how protein structures look, and it also uses other information like residue-residue correlations in large sequence alignments to infer which ones might be in contact. So when you turn off that option, it does not use specific templates, just all that other information. You may want to read the AlphaFold reference(s) cited in our AlphaFold help page: <https://rbvi.ucsf.edu/chimerax/docs/user/tools/alphafold.html>

As I understand it, even though AlphaFold can find many more templates than four for a given protein, it only uses the top four of those. See this previous response on our mailing list: <https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/...>

If you want to do homology (comparative) modeling based on one or more specific template structures, you could use the Modeller Comparative tool in ChimeraX instead: <https://rbvi.ucsf.edu/chimerax/docs/user/tools/modeller.html>

I hope this helps, Elaine ----- Elaine C. Meng, Ph.D. UCSF Chimera(X) team Resource for Biocomputing, Visualization, and Informatics Department of Pharmaceutical Chemistry University of California, San Francisco

...

On Feb 6, 2024, at 8:58 AM, Nadine Hellmold via ChimeraX-users <chimerax-users@cgl.ucsf.edu> wrote:

Hi dear support team, I am so happy to use AlphaFold directly in ChimeraX and it work's so smooth! Thanks a lot for it.

I was wondering what the feature "Use pdb templates when predicting structures" exactly means? I understood it like during the prediction up to 4 templates from pdb database can be used. I used the option for one prediction, for a protein complex prediction with five proteins, and found way more than four templates used for each protein.

If the option is switched off, what is used as a template then?

Bests Nadine -- Nadine Hellmold PhD student Department of Environmental Biotechnology | Helmholtz Centre for Environmental Research - UFZ

_______________________________________________ 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/