Thanks to all who replied. And thank you Ralf for your suggestion (quoted below), this worked very nicely! Somehow I missed this possibility in the online documentation. To reply to Pavel's question regarding how I figured out that tris is present at 1/2 occupancy: (1) density was first modelled with waters but they were too close together, and there was a fairly clear tetrahedral shape. (2) phosphate didn't refine well -- lead to huge negative difference peak at the P atom. (3) only other tetrahedral shape molecule in the soup was tris (TAM), and after refining just the tetrahedral core it was possible to see where the three arms extended. (4) b-factors for the tris were high relative surrounding protein atoms, and extraneous density suggested two alternate conformations for the nearby arg side-chain, one of which is too close to realistically exist next to the tris ligand. the other clue was that trial occupancy refinement of all atoms put most tris atoms in the 0.4-0.6 occupancy range. (5) tris with 0.5 occupancy, arg with two alt confs, plus a few waters at 0.5 occupancy is so far the best model with no significant different densities for this region. (6) the modelling will need to be confirmed with an SA omit ...

Hi Cathy,

...

The density suggests that when tris is not there, a side chain moves very close to that position

You could model this with alternative conformations. If you assign "altloc" A and B to two conformers of your protein, and just, say, B to your TAM, then the ligand would only "see" your B protein. If you don't define the B conformer (no atoms) then TAM wouldn't see that part of the protein at all. Actually, this approach seems better to me than working with atom selections.

Ralf