The PDB has quite a number of entries whose coordinates were derived using NMR. The most common situation is that the depositor provides some number of models plus the minimized average structure derived from those models. The PDB felt that it was most appropriate to create one entry for the average structure and to keep the multiple models together in another entry. This allows the user to view the multiple models readily. In order to allow the user to separate the models, the PDB created the following structure for the multiple model entries: header material MODEL 1 coordinates for model 1 ENDMDL MODEL 2 coordinates for model 2 ENDMDL ... CONECT MASTER END Note that the PDB numbers the atoms sequentially through the entire entry; i.e. atom numbering does not restart with 1 for each model. It would be most helpful if there were a way to handle this in mmCIF. The most compact way that I thought of would be to create new items that could appear in a list that would present the model number along with the first and last atom numbers (from _atom_site.id) associated with that model. But this assumes that _atom_site.id is the sequential atom number that appears in a PDB file. Alternatively an additional parameter could be defined in _atom_site that would present the NMR model number. This makes no assumptions about the contents of atom_site.id but means that an additional field must be present for each atom in a multiple model entry. In general, the PDB uses sequential model numbers starting with 1. But the largest NMR sets currently in the PDB contain more than 99999 atoms (current PDB limit given our format). In this case we create several entries. The first contains models 1 through N, the second N + 1 through 2N, etc. Because this issue relates directly to the representation of atomic coordinates, I feel that it would best be handled by the mmCIF committee. Thank you. Frances C. Bernstein