Our "refinement" consisted mainly of running undertaker with penalties
for breaks and clashes turned up high.  After each undertaker run, we
also did a quick optimization with gromacs (force field 0) and
repacked the sidechains with rosetta.  

Refinement 1 is try9-opt2, the best-scoring undertaker run with the
	the try12 costfcn.

Refinement 2 is try6-opt2.gromacs0.repack-nonPC, the result of the
	first undertaker refinement run after reoptimization with
	gromacs and rotamer optimization by rosetta.  It is rosetta's
	favorite of the models we built.

Refinement 3 is try10-opt2, which was optimized by undertaker from a
	chimera in which the loops
			S146-G151 
			P100-P108
			K20-N25
	were replaced by ones generated previously in our own attempts
	to predict T0308.  It did not score as well as the
	optimization directly from the supplied tr308.

Refinement 4 is try12-opt2.gromacs0.repack-nonPC, rosetta's favorite
	of the models built with the replaced loops.  Rosetta doesn't
	care much for it, and neither does undertaker.

Refinement 5 is try7-opt2, the reoptimization by undertaker of try6-opt2.
	It is yet another minor variant, submitted just to bring the
	number of submission up to 5.

None of these 'refinements' have really closed all the gaps and
resolved all the clashes, but they are about as good as we're going to
get in the time available, as we have never tuned our tools for doing
refinement.