I’m using Phenix to refine an EM-derived model that is a protein complex with a polyoxometallate (POM). The POM has the formula CoTiW11O40(8-), where the Ti and W (as bridged 6-coordinate oxometallates) form an cage around the Co(II) ion with overall Td symmetry. You can google “Keggin” and “POM" to get an idea of the shape. The published structure of the isolated POM modeled 11/12 occupancy of W at each oxometallate center, and 1/12 Ti. At 2.2 Å, our map shows no obvious placement for the Ti relative to the W centers. So far, for simplicity’s sake, we have used a ligand cif file that has 12 W centers in refinement (this was derived from the published crystal structure using Elbow). But to finalize the model, it seems important to acknowledge the presence of the Ti center. A few thoughts have come to us on how to proceed. We could simply annotate our model to indicate the presence of an unmodeled Ti center and continue with the current 12 W model or we could arbitrarily place a Ti ion in each POM (there are 4 in the complex). More realistically (?) we could place two POMs at each site - one containing nothing but W centers at 11/12 occupancy and one with nothing but Ti centers at 1/12 occupancy. But that generates two unrealistic ligands instead of one. The best thing to do would be to have a ligand cif file that has the atom occupancy built into it, but so far as I can tell, Phenix won’t accept that. I’ve also contacted the RCSB for advice, but would welcome any thoughts you all might have, especially if there’s a way to create an accurate ligand description. Thanks in advance, Arthur Glasfeld Mansoor Lab Oregon Health and Science University