James Holton
Department of Biochemistry and Biophysics, University of California
Physical Biosciences Department, Lawrence Berkeley National Laboratory
Abstract:
Shortly after the discovery of diffraction of x-rays from crystals in 1912, investigators such as Laue, Bragg, Ewald, Debye, Darwin and many others set to work quantifying the phenomenon. Nearly a century has past since these initial foundations, but the physics and mathematics derived at that time still hold to this day. What has changed is that textbooks covering these fundamentals are becoming harder and harder to find, and the equipment used to measure x-ray diffraction intensities has advanced through many generations. These advances, and the recent cryo-cooling revolution, have set new limits on what can and cannot be done with x-ray crystallography and perhaps obsolesced what were once optimal data collection strategies. Exactly what the new strategies should be can be established from the fundamental theory, once it has been updated to a modern context. For example, the word “pixel” was not used in scientific literature until 1965, more than half a century after Darwin derived a formula for the intensity of a spot.
In general, an optimal data collection strategy must strike a balance between data quality and radiation damage, and this requires the answer to several critical questions: How much exposure is required to solve the structure? How small can a crystal be before data collection will be a waste of time? How much redundancy (with shorter exposures) will add “too much” read-out noise? What about a better detector? What about a perfect detector? Answering these questions requires that damage, noise and signal be placed on a common, absolute scale. To this end, a quantitative simulator of the entire diffraction experiment was created and called "MLFSOM" (MOSFLM in reverse). The input to the simulator is a protein data bank (PDB) file and parameters such as photon flux, crystal size and detector performance characteristics entered in conventional units such as photons/s and millimeters. MLFSOM was used to produce images in SMV format that were subsequently processed with ELVES. The general result of these trials was that one and only one of the many sources of noise in the diffraction experiment will dominate a given data set, but the read-out noise of a modern detector cannot have a significant impact on anomalous data. The optimal strategy for MAD/SAD data collection was collecting a large number of very brief exposures, or “dose slicing”.
2009 Run
Sept. 23rd to Nov. 10th
