Hanspeter Winkler


Extraction of 3-D information from electron micrographs requires the application of complex 3-D reconstruction algorithms. In many instances new algorithms must be developed to deal with new reconstruction methods as well as improve existing ones or adapt them to new situations. My work involves development and implementation of new reconstruction algorithms.

One of these was a method for combining images of thin sections through a crystalline specimen that were cut at varying angles to different unit cell axes. The method, called oblique section 3-D reconstruction, is an adaptation of a method originally developed by R. A. Crowther. It allows the user to cut thin sections to reconstruct a thick object. The image to the right is the result of such a reconstruction. The sections used to make the reconstruction were ~150Å thick whereas the unit cell was 520Å thick.

We now concentrate on electron tomography, which is a reconstruction method that can produce a 3-D image of anything, whether crystalline or not. We have developed a new reconstruction method for processing tomographic data and are now applying methods to classify and average different motifs within a paracrystalline specimen. An outgrowth of this work was a method for 3-D distortion correction, which will not only correct for the distortion in the X-Y plane, as is typically done with crystalline specimens, but will correct for bending in the Z direction as well. Such a correction requires a tomogram of the paracrystalline specimen.

We have also applied our tomographic methods to non-crystalline specimens such as ice-embedded viruses and muscle proteins adsorbed on lipid monolayers (myosin V). For the latter we corrected the focus gradient in the micrographs prior to alignment with our marker-free method. Tomograms computed by weighted backprojection were further analyzed with methods common in single particle analysis of projection data (multireference alignment, multivariate statistical analysis and classification). We adapted these methods for 3D data in order to be able to characterize structural heterogeneity in out tomograms.