Our 3D ab-initio modeling procedure does not require nor does it make any prior model assumption. The input to our procedure are single particle images that have been picked from micrographs either manually or automatically. The main steps of our procedure are as follows:

  1. Class Averaging: classification of images with similar viewing directions, in-plane registration and averaging to improve their signal-to-noise ratio (SNR).
  2. Orientation Estimation: assigning orientations to the class averages based on their mutual common-lines.
  3. 3D Tomographic Inversion: reconstruction of an initial 3D model from the class averages based on their estimated orientations. 
  4. Iterative Refinement: alternating minimization procedure, that alternates between re-assigning orientations to all raw images based on the current 3D model, and 3D tomographic inversion for model updating. 


The algorithms employed in first and second steps (class averaging and orientation estimation) do not exist in other toolboxes and have been solely developed by our team (please refer to the publication list in order to read more about the algorithms). The third and fourth steps (3D tomographic inversion and iterative refinement) are classical and are implemented in many other toolboxes. We include them in our toolbox for the convenience of the user.