COMET

Image processing workflow in structural studies by electron tomography. A series of images is acquired from the sample at different tilt angles around a, typically, single axis. The images have to be mutually aligned. In high resolution structural studies, CTF has to be determined and its effects corrected for. Next, tomographic reconstruction combines the aligned images to yield the 3D volume or tomogram. Afterwards, the tomogram is typically subjected to denoising, to reduce noise with preservation of details. Segmentation then intends to decompose the tomogram into the structural components. Finally, if repetitive structures are present, they can be detected and extracted for further analysis. This analysis typically includes 3D alignment and averaging of the subtomograms to obtain a high resolution map. Also, quantitative analysis of the distribution of the subvolumes within the tomograms (a.k.a. visual proteomics) are often done. 

Illustration of some of the methods that we have developed. Fast iterative tomographic reconstruction yields tomograms with better contrast (Vaccinia virion shown here). Denoising with sophisticated methods allows noise reduction with preservation of the structural features (HIV-1 virion shown here). We have also developed methods for automated segmentation of tomograms, which greatly facilitate interpretation in an objective fashion (Golgi complex and mitochondrion shown here). We have also implemented and used techniques for subtomogram analysis. Bottom panel:  (a) A tomogram of a thin cryo-section of yeast where (b) the ribosomes were automatically detected using, as a template, a previous ribosome density map obtained by single particle cryoEM. (c) The extracted ribosomes were mutually aligned and an average was computed. (d) Finally, the ribosomes were put back into the 3D space in their location and with the determined orientations, thus building a ribosomal atlas.