Virtual atoms remain superimposed on C-alpha

[owen-r-rogers]

Hello,

I've noticed that in a fair amount of structures output by rfd3 (for motif scaffolding) there are residues where it looks like the virtual atoms didn't get removed, resulting in very badly-scoring files (cart_bonded ~ 10,000 REU). In these cases it looks like a glycine in pymol, but the .pdb file will list atoms CB onward as having the same xyz coordinates. Is there a way to cleanup virtual atoms/filter for residues where there are huge steric clashes from virtual atoms remaining on top of the C-alpha?

Thanks!

[Ubiquinone-dot]

It is pretty curious to me that this is happening - can you provide more details of your exact inputs? You can use read_sequence_from_sequence_head: False which should mean that the sequence assignment is based on the generated structure over the models' own prediction but we haven't found it to be unreliable in general