Thresholds in Protein Crystallography: Finding the Limits of Conventional Techniques to Inspire Unconventional Thinking
James Holton
Lawrence Berkeley National Laboratory
Protein crystallography has advanced considerably since the turn of the century, but it is still a difficult technique. The speed of successful structure determination has increased considerably, dropping from months to minutes in some cases. But the unsuccessful structure determinations deserve attention as well. The most "interesting" projects always seem to be the ones that push the boundaries of our best methods to the "threshold" where solving the structure transitions from being merely difficult to being impossible. It is the job of methods developers to continue to push this "threshold" back. So, it is logical to begin by establishing where these thresholds are. What is the limit to how many heavy atom sites you need to solve a MAD structure? How different can a search model be before molecular replacement won't work? What is the limit to how small a protein crystal can be? How much radiation damage is too much? How can you tell? How many projects fall into the "gray zone" between data sets that are trivial to solve by standard methods and data sets that cannot be solved with the most advanced algorithms and cleverest tricks? Is it possible that even more advanced techniques can push these limits further? If so, which ones will have the most impact?
We have investigated these questions by examining the common problems encountered by PX beamline users and testing the limits of algorithms with realistic simulated data. The answer to all these questions seems to be signal-to-noise. The ratio of the signal of interest to the error associated with it must be greater than approximately unity for the structure solution to work. The "gray zone" appears to be rather narrow, indicating that current algorithms are actually quite effective (when used correctly). The most impact can now be had by improving data quality and new experimental techniques to push the limits of structure determination further in the future.
