Hi Troy, if the C and N are within 3 A the peptide link should be added automatically by phenix.refine, and the distance should be kept close to the ideal distance at all times during the refinement. If that's not the case, something is not right. You can check in the .geo file if phenix.refine generated the peptide link. For example: bond simple: (4, 8) " C ASP A 443 " segID="A " " N SER A 444 " segID="A " distance_model: 1.32927 distance_ideal: 1.329 weight: 5102.04 If the link is missing because the atoms are too far apart, you can redefine pdb_interpretation.link_distance_cutoff=4 or some larger value, although it is probably better to regularize the model in coot as suggested by Chris. But once the link is there, it should be stable during refinement. Let us know if that's not the case. Ralf