PSI Structural Biology Knowledgebase

PSI | Structural Biology Knowledgebase
Header Icons
E-Collection

Related Theme

One from many

SBKB [doi:10.1038/sbkb.2011.34]
Technical Highlight - August 2011
Short description: Xsolve utilizes many different approaches to streamline and enhance automated structure determination.

Schematic of the multiple computational approaches combined by Xsolve to automate structure determination.

One of the biggest hurdles to high-throughput X-ray crystallographic structure determination is the time and manpower required to move data from the collection stage all the way through to a refined structural model. In theory, automated approaches can assist by cutting down the time researchers need to spend on the earlier stages of structure determination and maximizing the quality of the initial structural model. However, for a given data set it is hard to say which combination of programs and parameters will return the best solution for the data set.

A major focus of the Joint Center for Structural Genomics (PSI JCSG) has been the development of improved tools and approaches for automating the determination of protein structures. Now, van den Bedem and colleagues present a highly effective tool for just that purpose: the Xsolve crystallographic software pipeline. Capable of handling every step from reading the diffraction image to producing a partially refined three-dimensional model, Xsolve sets itself apart by not trying to pick the best approach for structure determination. Instead, it runs multiple third-party programs for identical tasks while also exploring key parameters independently and in parallel at all stages of the structure determination process to determine the best structural model, which is then refined by a crystallographer.

To date, almost 90% of the JCSG's 770 MAD and SAD structures have been solved using Xsolve. In an analysis of 36 structures determined using Xsolve, various programs were needed to solve all structures, indicating the utility of the Xsolve approach, which allowed the strengths of each program to shine while balancing their weaknesses. For example, although Buccaneer yielded the most complete trace for 33 structures, it had more errors than other programs, whereas ARP/wARP gave the best degree of completeness at the highest resolutions. The software program ConsensusModeler, developed by the authors and run after the models are generated, then combines different traces from separate approaches into a final structure with a higher degree of completeness and fewer errors than would be available by running any individual program alone.

Xsolve does, however, require a fair bit of computational time and power, but that cost is more than made up for by the quality of the results and improved efficiency. Thus, Xsolve stands to become a powerful tool in the quest for high-throughput structural biology with minimal additional cost.

Steve Mason

References

  1. H. van den Bedem et al. Distributed structure determination at the JCSG.
    Acta Crystallogr D Biol Crystallogr. 67, 368-375 (2011). doi:10.1107/S0907444910039934

Structural Biology Knowledgebase ISSN: 1758-1338
Funded by a grant from the National Institute of General Medical Sciences of the National Institutes of Health