> > No disorder is involved. > > > > The occupancy of an (fully occupied) atom on an n-fold rotation axis is > 1/n > > If a two-fold, 1/2 > > If a three-fold, 1/3 > > > > When you sum over all the atoms in the unit cell, application of the > symmetry operations to atoms lying on the rotation axis generates atoms with > unchanged coordinates. Hence to generate a fully occupied atom on a n-fold > symmetry axis, the original occupancy has to be 1/n. > > >
I guess I have a few small points to add: 1) If a site on special position is not fully occupied, then the occupancy is q/n, where q is actual occupancy (for fully occupied site q=1). Since ATOM record in PDB file does not (directly) tell you whether the atom is on special position or not then it is not straightforward to know whether the occupancy of atom in question is not 1 due to disorder or due to symmetry. 2) Let's consider a hypothetical case... The maximal multiplicity of a spacial position in proteins is 24 (is that right? please correct if not), and the precision of the occupancy field in PDB file is only two digits, like: 1.00. Imagine you have a fair amount of very heavy atoms all seating at special positions with the multiplicity 24, and your structure is relatively small. Then what you will be forced to put in the PDB file is: 1/24 ~ 0.04 and not 0.042. I'm not sure if this rounding error is significant or not (and what is "significant"), but just to keep in mind another source of mismatch between reported and recomputed R-factors. 3) If I add an atom onto special position using Coot, will it set the occupancy as 1/n automatically or do I have to to edit the PDB file myself (and remember to do so)? (another room for an error) So, having said this, I tend to think that using occupancy 1 (and not 1/n) is a good thing. Pavel.
