We had good alignments for both domains of this model, but were
uncertain how the ubiquitin-like domain and the HAUSP domain would
interact in the crystal.  The most likely possibility seemed to be
that the ubiquitin-like domain would be in the binding pocket of the
HAUSP domain---probably domain-swapped in a multimer.  Since we lacked
tools for optimizing multimers, we elected to dock the domains as a
monomeric protein.

We superimposed try3-opt2 on 1nbfA and 1nbfD, then cut the domains
apart where the linking region would require the least rearrangment.
After reoptimizing, we got try5-opt2, which is our best-scoring model
with an unconstrained cost function, and is submitted as model 1.