I noticed that the CONNECT records of ligands generated in elbow get deleted in phenix.refine. Even specifing both the protein and ligand file (with CONNECT records) in the .def files, does not retain these records. For us they are coming in handy since they specify the bond order of the ligands. This is important for display in PyMol, which assigns bond orders based on these records.
Would it be possible to keep the CONNECT records in the output file of phenix.refine? Obviously the numbering would need to be adjusted and I have not yet checked if some of the model building programs like coot would be able to handle this situation correctly.
Yes, that's exactly why maintaining the CONECT records is a real project. As an aside, the PyMOL trick of repeating records to convey the bond order isn't used anywhere else as far as I know (but I like the idea). I'm in the middle of re-doing the fundamentals of how phenix.refine handles pdb files. Once that's settled we will review how much effort it would take to maintain CONECT records. Ralf