Obtaining initial reconstructions in three-dimensional electron microscopy (3D EM)

Imaging macromolecules in a large variety of states — conformations, intermediates in a reaction pathway, etc. — in a single sample, under physiological conditions and at high resolution is arguably cryo-EM’s most exciting promise. Achieving this goal will require collecting and analyzing data sets of a size incommensurate with the current, highly interactive manner in which structures are obtained. There are three steps in determining a structure by cryo-EM: (1) collecting data; (2) obtaining the initial, low resolution structure and (3) refining it to high resolution. Efforts to automate steps (1) and (3) have been under way for some time. Obtaining the initial structure, however, remains a bottleneck for any new project. The two traditional approaches used to achieve this—Random Conical Tilt (RCT) and Angular Reconstitution (AR)—both suffer from shortcomings that make them unsuitable for full automation. The former requires significant intervention by the user while the latter has as a major underlying assumption the absence of heterogeneity from the sample. We have developed a novel method—Orthogonal Tilt Reconstruction (OTR)—that combines the strengths of the two traditional methods and avoids their shortcomings. This makes OTR amenable to full automation.

The specialized data collection required for OTR has been automated by the groups of Bridget Carragher and Clint Potter at Scripps as part of their Leginon automated data collection system. However, data processing for OTR is currently carried out with available software packages that require a level of expertise that places the technique beyond the reach of most potential users. We are interested in creating software that will interface between the user and these existing data-processing packages to significantly lower the level of expertise required to apply OTR. We would also like to include features that will help in dealing with some of the difficulties encountered when obtaining initial reconstructions, such as flattening and heterogeneity. Combining this software with the automated data collection would bring us much closer to the day where cryo-EM becomes a routine structural biology technique.