Choosing the redesigned structure

[barthelemymp]

Hello ProteinMPNN team,

Thank you for your work!
In your paper you refold some native protein and compare with alphafold.
("We reasoned that ProteinMPNN might generate sequences for native backbones more strongly encoding the structures than the original native sequences, as evolution in most cases does not optimize for stability, and completely redesigned a set of 396 native structures").

1. If I understood correctly, the refolding part is sequence only (No template, no MSA), which sounds like the fairest way to use alphafold in your case. Am I right there ?
2. How have you chosen this 396 structures ( and can you share the list). As for the few proteins I tried randomly so far, alphafold with no template and no MSA is just too wrong to be uswed as an assessment. Whereas in your plot a large part of 396 native proteins seems to have reasonable LDDT after refolding.

Best

Barthelemy