TB Structural Genomics Consortium (TBSGC)

Structural genomics often brings us three-dimensional structures of proteins with no known functions.  To shed light on such structures, Debnath Pal and David Eisenberg at UCLA have developed ProKnow (http://www.doe-mbi.ucla.edu/~dpal/ProKnow) which annotates proteins with Gene Ontology functional terms.  The method does so by extracting features (such as 3D fold, sequence, motifs, and functional linkages) from the protein structures, and relates these to probable functions.

 

A combined biochemical and informatics approach has been applied by Michael Strong, Shuishu Wang, and David Eisenberg at UCLA to discover the binding partners for two proteins and to produce a soluble complex for crystallographic structure determination.  Neither member of the complex expresses well or is soluble on its own; only when they are co-expressed is the hetero dimer soluble and crystallizable.  This complex is formed from a PE protein and a PPE protein encoded by the genome of M. tuberculosis.  The method, which starts with finding a pair of proteins having a strong indication of functional linkage, should be general for finding binding partners for numerous proteins that are not soluble on their own.

 

 

 

CrysFind, developed at Lawrence Livermore National Laboratory, has been benchmarked on test images acquired on a Velocity11 Versascan instrument and initial tests comparing results from a human observer show a disparity of 1% for finding crystals (or 1% false positive) and 3% for observing no crystal (or 3% false negative) for crystals with the smallest dimension >100 µM. CrysFind, in association robotic imaging and an SQL database enables walk away automation for tracking and scoring of crystallization experiments. In association with commercially available high magnification crystallization imaging systems CrysFind could enable auto detection of crystals <20 µM.

 

 

 

Lawrence Livermore National Laboratory members have automated coarse screening by adapted robotic liquid handling to stochastic sampling of a very broad crystallization sample space. This is accomplished by on the fly combinatorial mixing of cocktails from stock reagents known to be useful in crystallization experiments. The implantation of automated course screening enables the demonstrably most efficient sampling of crystallization parameter space as well as a much more thorough search than commercially available screens. The same equipment and set of reagents is used for fine screen optimization.

 

 

 

Lawrence Livermore National Laboratory members have co-developed (with Robbins Instruments) labware, the Intelli-Plate, for sitting drop 96-well crystallization with sub-microliter drop volumes. The Intelli-Plate enables automated crystallization without the need for customized robotics. The group has also co-developed (with Apogent Discoveries) a high precision 96 well liquid dispenser (the HydraPlus1) for the automated setup of crystallization experiments in 96 well formats with sub-microliter drop volumes. The group is co-developing (with Innovadyne technologies) the next generation liquid handling robot, Screen Maker 96+8, which will enable high precision nanoliter volume crystallization with stochastic sampling.

 

 

 

Elves is an expert system developed in the Alber lab at UC-Berkeley to automate available crystallographic software. Recent improvements include robust data processing and automated molecular replacement. Development has begun on Collector Elves to begin to automate beamline operation.

 

 

 

James Berger at UC-Berkeley and Steve Quake at Caltech invented a small-volume crystallization chip to screen conditions efficiently and speed crystal growth. The technology was commercialized by Fluidigm. It is widely used by crystallographers worldwide.

 

 

 

Numerous innovations in beamline design at ALS Beamline 8.3.1 were exploited by TBSGC scientists in order to efficiently collect and process data. (See MacDowell, A.A. et al. (2004). Suite of three protein crystallography beamlines with single superconducting bend magnet as the source. J Synch. Rad., submitted.)

 

 

 

Geoffrey Waldo and his colleagues at Los Alamos have developed systems for evaluating whether a protein will fold properly when expressed in cells or in vitro, and for identifying whether a protein is present in a solution in a soluble form. One method is based on fusions of the protein with reporter proteins such as green fluorescent protein, and another is based on complementation of a green fluorescent protein lacking one strand by a tag fused to the end of the protein. These methods can be used to engineer proteins to improve their folding and solubility.

 

 

 

Tom Terwilliger and colleagues at Los Alamos, working with the PHENIX consortium (http://www.phenix-online.org) have developed the SOLVE/RESOLVE software suites (http://solve.lanl.gov) for carrying out fully automated determination of 3-dimensional structures of proteins once X-ray anomalous diffraction data have been collected.