Disconnected backbone?

[kimdn]

I just simply ran default input file/option of design.sh and obtained

I am just curious whether "disconnected" peptide backbones are ok as above.
With my experiences with pymol, chimeraX, I assume that distances between these designed amino acids are too "far".

When I generated cyclic peptide backbones by RFdiffusion's RFPeptides, I didn't observe "disconnected" backbone as

(of course, I observe 'small' disconnection between N-terminal and C-terminal, but this small disconnection was also observed even with native/natural cyclic peptide)

I read

#The design process can result in clashes - especially in the cyclic case.
--
#We recommend running relax, but don't provide this for modified residues

in design.sh

Maybe relaxing by AlphaFold2's later stage/amber/classic Rosetta's FastRelax will connect these "disconnected" backbones?

Many thanks.

[kimdn]

Just FYI, I generated this kind backbone disconnected protein when I added non-canonical amino acid by RosettaFold All Atom (RFAA) as well.

Author there suggested to re-run the RFAA simply because it has stochastic nature.

[patrickbryant1]

Low confidence designs may have disconnected residues. Relax may fix this, but I don't recommend that if the plddt is low since the prediction is likely incorrect.