IIMS Workshop Report

Overview Summary

Structural electron microscopy has matured over the past decade. Much of the recent progress results from two, complementary, advances. The first was the introduction of cryo-electron microscopy that preserves native structures in vitrified water. The second was the development of image processing techniques that allowed the determination of three-dimensional structure from the relatively low-contrast image generated by cryo-electron micrographs of unstained specimens. These advances make electron microscopy an important complement to NMR and X-ray diffraction. An increasing number of projects use this complementarity directly. Often a three-dimensional structure derived from cryo-electron microscopy (cEM) of a large complex is used to generate an atomic models in combination with structures of components determined by X-ray crystallography or NMR. The transfer of information is not only from high resolution to low. Some X-ray crystallographic structures have been solved by using models derived from microscopy to determine phases.

This recent progress shows the power and desirability of sharing the information from microscopy within the structural community. This was not routinely possible until recently. An X-ray crystallographic or NMR experiment produces atomic coordinates that are routinely deposited in the Protein Data Bank (PDB) along with supporting data. A cEM experiment produces a map showing the potential distribution within the object. A few cEM studies produce atomic models however many, particularly as lower resolution, only produce maps. Deposition in the PDB requires coordinates. The only option for depositing the map was the indirect one of depositing structure factors calculated from it along with the coordinates of the atomic model. A number of cEM structures have been deposited this way. They remain the exception rather than the rule. Most lack the coordinates needed for PDB deposition. Only a handful of the hundred or so EM structures determined each year are available in the PDB. This situation limits the usefulness of microscopy as a structural technique.

A barrier to the routine deposition of EM structures is the lack of agreement on an image or map format. There are at least a dozen image and file formats in common use and there has been little agreement on standardising on a single one. A meeting was held at the European Bioinformatics Institute near Cambridge to try to explore solutions to this problem and others that limit the deposition of EM structures. The Wellcome Trust supported meeting was held under the auspices of the Information Interchange for Molecular Structures (IIMS) initiative of the European Union. It assembled representatives of many of the active labs in software development for electron microscopy. This workshop was also an opportunity to acquaint the community with the progress that has been made in establishing a database for electron microscopy maps.

The current EM database (EMD) profits from the experience with BioImage, a previous Commission funded project that established a limited collection of light and electron microscopy datasets. The development of the EM database was funded under the Fifth Framework Programme of the European Commission. The effort includes the definition of flexible metadescriptors for electron microscopy entries, a deposition interface and a search interface for the entries. The deposition interface (emdep) avoids the difficulties of the lack of a standard file format by converting a deposition in any common format to the Collaborative Computing Project Number 4 (CCP4) format, the most common map format used by protein crystallographers. This choice is particularly appropriate because one of the aims of the database is to enhance the interaction between microscopy and other structural techniques. The database in linked to the Macromolecular Structure Database (MSD) at the EBI. It should share the same permanance as the PDB. The intention is to link it to the other protein structural databases.

The data provided would be far more valuable if it were accompanied by criteria of reliability that would aid the interpreter. There is not a single common standard that covers the entire range of possible entries. There are five types of data that form the bulk of electron microscopy structures: electron crystallographic, symmetric single particles such as icosahedral, helical structures, single particles without symmetry, and tomography. The criteria for reliability in each of these techniques vary. Crystallographic and helical reconstructions can use the measured spatial resolution of their data and such indicators as phase error as criteria. The reproducibility of the reconstruction, as assessed by the plot of the Fourier Shell correlation (FSC) between two independent reconstructions, is often used for single particle reconstruction. The participants in the workshop agreed that a single resolution number would be less useful that the curve showing the FSC as a function of special frequency. It was agreed that this curve should be deposited with single particle reconstructions along with any mask used for its calculation. Tomographic reconstructions are derived from multiple tilted views of a single object and hence the fineness of the sampling is used as a criterion for reliability. The establishment of a database will ease the development of new measures of reliability.

The workshop discussed many of the issues that arise from deposition and the presentation of the data in a form that will be useful to non-specialists. Initially the database will implement immediate release of the header information characterising the entries but provide hold periods ranging from immediate release to four years for the maps and structure factors. This may seem at odds with the, now standard, policy within crystallography of relatively rapid release. This reticence is perhaps easier to understand when one remembers that microscopy is still a relatively new structural technique. As the drive for high throughput methods spreads through microscopy, one can expect the speed of release will increase as it has in other structural techniques. This first step is important because it establishes the database as a resource for the community.