ERROR Report: Assertion `res2.is_bonded(res1)` failed causing crash

[wang-zs1]

Hi,I ran RFdiffusion and ProteinMPNN according to the parameters in Article "Accurate de novo design of high-affinity protein-binding macrocycles using deep learning", and all these results worked well. However, when executing a protein relaxation (FastRelax) and cyclization (PeptideCyclizeMover) workflow using PyRosetta with RosettaScripts, an assertion error Assertion res2.is_bonded(res1) failed occurs, leading to program crash. The error happens during the energy calculation phase and may be related to abnormal residue connection relationships. I don't know if this is a normal phenomenon.The error report is as follows:

core.conformation.Residue: [ WARNING ] While updating residue 16's connections to residue 17, a connection to residue 1 was overwritten!
core.chemical.AtomICoor: [ WARNING ] IcoorAtomID::atom_id(): Cannot get atom_id for POLYMER_LOWER of residue LEU 18. Returning BOGUS ID instead.
core.conformation.Residue: [ WARNING ] missing an atom: 18 H that depends on a nonexistent polymer connection!
core.conformation.Residue: [ WARNING ] --> generating it using idealized coordinates.
ERROR: Assertion res2.is_bonded(res1) failed.
ERROR:: Exit from: /home/wzs/wzs/rosetta/source/src/core/scoring/etable/count_pair/CountPairFactory.cc line: 220

[roccomoretti]

It's a bit hard to say without seeing your input PDB, but it sounds like you might be running up against the same issue as is seen in PR #150 (note that there's some issues with the implementation of that PR, which is why it hasn't been merged yet).

One work-around would be to switch the order of cyclic peptide and protein -- I think you can sidestep the issue by putting the cyclic peptide first. Another would be to add some sort of non-polymeric residue (e.g. a ligand or a metal ion) between the two chains, which I think (but haven't tested fully) would cause Rosetta to know to not merge the two chains.