3D Reconstruction

Access: CP::3D Model::Reconstruction (Tab)

The Reconstruction module provides functions for image angular reconstitution, Random-Conical-Tilt (RCT) reconstruction, and model refinement.

Reconstruction & Refinement Parameters

PARTICLE provides the following methods for 3D density map reconstruction and refinement includes:

• ab initio : use the common-line principle to determine the relative angular orientations of a set of images, from which to reconstruct a density map.  High SNR class-averages should be used in order to achieve a reliable alignment.  The Symmetry property will be utilized in the process of both angular reconstitution and model reconstruction.  The reconstructed models will be sequentially named as "$mapname_AB#.mrc", where "#" = {1, 2, ... }.

• CQ Refine : after an initial reconstruction (e.g., by the ab initio method), identify one or a few exceptionally good classes at distinctive views and refine one density model for each selected seed.  The refined models will be sequentially named as "$mapname_CQ#.mrc", where "#" = {1, 2, ... , number of seeds}.

• PM Refine : model refinement via the conventional projection matching.  A reference model should have been loaded into the map editor CP::Editor::MAP.  The refined models will be sequentially named as "$mapname_PM#.mrc", where "#" = {1, 2, ... }.

• MC Refine : model refinement via Monte Carlo simulation for optimization.  A reference model should be present in the map editor.  The refined models will be sequentially named as "$mapname_MC#.mrc", where "#" = {1, 2, ... }.

• R.C.T. : random-conical-tilt model reconstruction (see below).

• Build Map : 3D reconstruction from a stack of aligned particle images.

And the parameters common to most of the methods are:

• Symmetry : the symmetry of the 3D model (only the C point-group has been implemented at the moment).
• T-Search : the off-centering tolerance in translational search (in pixel).
• R-Search : the angular step in 3D rotational search (in degree).
• Filtering : Gaussian filtering to reduce noise in the images.
• Masking Radius : a mask that encircles the particles for reconstruction.
• Symmetry Weight : a symmetry-related weighting used in common-line-based angular reconstitution.
• Correlation Threshold : a quality threshold above which the images will be included in the model reconstruction.

In addition, there are a few method-specific parameters to be defined in the Refinement Parameters group:

• AB Models : number of models to be constructed by the ab initio method.
• AB/CQ Attemps : number of attempts in each AB/CQ reconstruction.
• MC Iterations : number of cycles in the MC refinement.
• MC Restarts : number of restarts in each MC iteration.
• PM Iterations : number of cycles in the PM refinement.

Activate the Reconstruct button will initiate the process of model reconstruction and/or refinement.

RCT Reconstruction

The RCT method uses particle tilt-pairs for 3D model reconstruction.  Because the tilt geometry comes from experimental observation, this method can significantly reduce potential error in particle alignment and the initial model reconstruction.  Moreover, the geometric restraints from tilt-pair data provide an unbiased mechanism for model quality evaluation.  The importance of the RCT method merits a section of its own in the documentation.