I may have missed it, but I can't seem to find an equivalent to the PDB MTRIX entries for non-crystallographic symmetry in cifdic.m95. If it is not already there, I would suggest an addition to the atom_sites category similar to the *_transf_matrix[i][j] and *_transf_matrix[i] which now have *=cartn and *=frac, with a new *=non_cryst, and a linked non_cryst_transf_id to allow for multiple transformations. Alternatively, I suppose one could add it to the struct_asym category as Phil Bourne has suggested in pdb2cif, but putting it with atom_sites seems "more natural" (whatever that means -- there's no accounting for taste, certainly not for mine). In any case, my apologies to one and all if I missed something that has already been taken care of. -- H. J. Bernstein