I submitted a job to the ROSIE antibody queue, which ran successfully, but all the result files now have a discontinuity in the backbone 2 residues before the start of CDR L2. No residues are missing. I ran it again without the H3 modeling, and the discontinuity isn't there, so it's something during the high resolution modeling that is causing a problem. Is there anything I can do to prevent this from happening? Should I use the light chain structure from the grafted model as correct, and use the heavy chains from the more rigorous protocol?
I have attached an example PDB which contains the discontinuity.
The best I can tell, it looks like there might be an issue with detemining the position of the CDR loops. If it's getting confused as to where the loops are, that might mean that it puts a cutpoint in the wrong place, leading to things opening up.
Do you still have the links to the runs (both with and without the H3 remodeling)? Having more details about how the run progressed would be helpful in tracking this down.