search

google-deepmind / alphafold

Open source code for AlphaFold 2.
Apache License 2.0
12.92k stars 2.29k forks source link

Using rotamers for mutations #943

Open n1812 opened 5 months ago

n1812 commented 5 months ago

I was mutating a structure using the Rotamer tool on ChimeraX. I mutated a single amino acid (Tyrosine) to a Proline using the Richardson Common Atom library. However, when I ran AlphaFold on it, the predicted 'best model' structure they gave me had a different amino acid (Leucine) on the position I previously mutated the Tyrosine to the Proline. I was just wondering if I did something wrong or if AlphaFold works differently than I thought.

tomgoddard commented 5 months ago

Are you sure you ran an AlphaFold prediction? I think it is more likely you fetched a close matching structure from the AlphaFold database which did not have the same sequence. That is what the Fetch button does on the ChimeraX AlphaFold panel. If you ran a prediction using the Predict button in ChimeraX then you can see exactly what sequence it used in the AlphaFold Run panel that is shown during the computation. There is no way AlphaFold would put leucine where you had a tyrosine or proline specified. Another possible problem is the AlphaFold model always has residue numbers starting at 1 and that probably does not match the residue numbering in your experimental structure.

You would be better off discussing this issue on the ChimeraX mailing list (chimerax-users@cgl.ucsf.edu) since it is certainly not an AlphaFold issue.