Diffraction Methods Gordon Conference 2002 Notes

Conference Facilities

Location: Connecticut College, New London, CT. This location sucks. The venue is too large (2 Gordon Conferences + a High School Teacher conference does not make for an intimate setting), and New London appears to be a fairly depressed area. Don't even start me on the dorms.

Anyway, when not whining about the conference infrastructure, here are things that caught my eye at the conference:

MAD and SAD

A fair amount of emphasis was placed on single-wavelength anomalous diffraction (SAD) at this conference, in particular by B.C. Wang who touted possibilities of high-resolution SAD using home sources and the weak signal from sulfur. (Wayne Hendrickson was politely skeptical about the revolutionary possibilities of this method). Nevertheless from conversations it became apparent that iodine/iodide SAD data has been useful in some cases, either as NaI (Z. Dauter's method) or KI/I2 (an old method revisited by Gwyndaf Evans). Certainly proponents of Maximum Likelihood Phasing techniques (Randy Read, Gerard Bricogne) seemed very attracted to SAD as a technique. I have solved one structure using SAD, although that was because the MAD data seemed to produce worse maps than a single wavelength.

Nevertheless MAD is still the major phasing method. Wayne Hendrickson gave a talk where he expounded the possibilities of less-traditional elements for MAD phasing, but he stuck to his guns on what he considers the best data collection strategy: inverse beam. Others have observed that very high redundancy can maximize the anomalous signal in SAD cases: 360-700 degrees often produces a better signal/noise in anomalous data than more conservative data collection ranges. However Wayne pointed out that collecting such large amounts of data may be better for SAD, but may jeapordize the MAD experiment (likewise the inverse beam, move-between-wavelengths method jeapordizes the SAD experiment) because of radiation damage.

Zhen-Quing Fu from B.C. Wang's lab reiterated (endlessly) a technique advocated by others: that the "anomalous" signal between centric reflections is an estimate of the error in the measured anomalous signal since centric reflections have no anomalous difference between I+ and I-. This method is also on Bernhard Rupp's website and has been mentioned on the CCP4 bulletin board in the past. (Phil plans to implement this out of Scalepack in the near future).

Membrane Protein Crystallisation

Automation and other Hardware

There were some fairly standard reiterative talks (e.g. Steve Muchmore, B.C. Wang) about the possibilities of beamline automation and how it will magically speed up the screening of heavy atom derivatives. BNL's X12 beamline (via Howard Robinson) seems to have the most elaborate throughput plans at this point, with long term storage and retreival of samples, overnight rapid screening passes through crystals, slightly longer wedges of data from potential heavy atom derivatives etc, with a web summary of results for the experimenter to select the next experiment in the series. The ultimate in armchair/FedEx crystallography. There's convergence in the types of technology in use but no actual convergence in the formats yet. The field appears relatively open.

A small technical note: beamline automation has settled on Hampton style mounts as the de facto standard. No consensus as to pin length has emerged.

Nevertheless I feel quite strongly that this has a great deal of potential for somewhere like X9A, where a large number of prefrozen crystals could be screened automatically overnight, with a review of results the following morning followed by more extensive data collection on promising samples.

There were some talks on new detector technologies, including a CCD detector with a lens as the optical element (potentially better than the fiber optic tapers), this particular design did not seem to offer any actual benefits over existing fiber designs. Ed Westbrook had a poster on prototype technology for digital pixel counter detectors (generally believed to be the eventual state of the art sometime in the future, but not now). Ed Westbrook doesn't have the best track record in delivering technology, however. The other talk presented some data on the large flat amorphous Se detector in development by MAR which may in fact have better performance than a CCD (in particular a smaller point spread function), but is still in the prototype stage. This detector was also touted at least year's ACA meeting in Los Angeles, so development of it seems to be taking a while.

Other Ideas

A few talks provided examples of phase extension from low resolution cryo-EM maps (virus structures and the ribosome), use of heavy atom clusters (ribosome) and Ir(NH3)6 and Os(NH3)6 as useful heavy atom derivatives. There were no instant solutions, however. It was reiterated that the heavy atom clusters normally provide very little phasing information beyond 6 Angstrom and nobody seems to have routinely got around that problem (disordered sites and unresolvable heavy atom substructures are the cause of this).

Python and Structure/Visualisation

Python seems to be emerging as a useful scripting language with several toolkits in various stages of development. The PHENIX project plans to build an entire semi-automated customisable graphically-driven crystallography suite for these high-throughput times out of it. Michel Sanner presented some glimpses into what seemed quite an impressive macromolecular visualisation toolkit also in Python. The program Pymol has also been around for a while now, and may be worth a closer look. Other groups (e.g. CCP43D) also seem to be using Python.

Other Software

Just a couple of modules for CNS from Michael Chapman's lab at Florida State University and some exotic visualization software from Austin.