Perhaps inspired by the fierce debate on this BB and elsewhere about how to model disordered side chains, I would like to present the following scenario (my current refinement project): The structure was solved by molecular replacement, is 3.25Å resolution, and ~70% solvent. I have 6 copies of a large complex in the AU (~100kDa per protomer). Only one copy is fully ordered. In the other 5 copies, a entire domain (~25kDa) is largely disordered (some have patchy residual density, but not readily interpretable). Altogether, this means that I am missing ~20% of the total protein in the AU. It seems that how you model this much "missing" material could have a significant effect on the final refined model. This is something we have observed a number of times, so I am wondering if anyone can suggest ways to deal with this aside from just leaving the domains out entirely. In this case, the missing domain is connected by two short hinges that restrict the rotational degrees of freedom considerably (you can think of the missing domains as a door--it can only pivot about the hinge, can't be flipped upside-down or sideways, etc., and is anchored close to the door frame). Additionally, crystal packing and steric constraints restrict the "sweep" of the missing domains to further pinpoint their location. Is it worth trying to model this? My impression from some earlier posts was that most people were content for disordered regions to be modeled as bulk solvent (rather than fiddling with masks to also cover the expected location of protein, etc.), but I wonder if you may actually substantially improve the model when this much protein is disordered. I guess I am imagining something analogous to a rigid body fit to the mean position of the disordered domains and a TLS-like ADP description of the motion (with a very large magnitude to account for the large displacement), but I am open to suggestions. On a final note, regarding those pesky missing side chains: any thoughts on trying to employ a "Ringer"-like approach to model some of these (Fraser, et al., Nature 2009, 462(7273):669-673)? Is this practical (maybe this would add to many additional parameters)? Best, Damian Ekiert