Success Rates in Protein Crystallography
    Click for Graph
  1. Obtain milligrams of pure, soluble protein.
    1. About 60% of sequences express well.
    2. Only one half (prokaryotic) to one quarter (eukaryotic) of these are soluble.

  2. Obtain high quality crystals.
    1. Typically hundreds of crystallization conditions are tested.
    2. Crystals must be singular, sufficiently large, and preferably neither needles nor plates.
    3. Crystals with large numbers of copies in the asymmetric unit are problematic.
    4. About half of highly expressed soluble proteins crystallize, but only about one third of these crystals are suitable.
Overall success rates (Thornton): By May 2012, of ~200,000 targets cloned in Structural Genomics Projects, 11,000 had been solved (5.5%). Of these, ~5,800 nonredundant solved targets represent 36% of the (2001 likely underestimated) ~16,000-target goal of Structural Genomics.

by Eric Martz, University of Massachusetts, July 2003 (revised February 2004, July 2005, June 2006, April 2007, May 2012)
Thanks to Byron Rubin for some insights here.
Further reading: